Protechsoft

Supply Chain Management

Introduction

Discerning supply chain management as an exquisite part of stupendous growth is one of the most paramount realizations that any firm must make. The true question stands, why exactly do we have to believe that the supply chain management framework is not secondary to a colossal growth strategy but works simultaneously with the multitude of strategies that are embedded in the strategic growth framework? To understand that, lets first understand what a supply chain is exactly:

Definition of Supply Chain

A supply chain is a network of nodes, each that is assigned a responsibility which chooses itself from procurement, transformation or transmission of materials and information all the way from the ultimate transmitter (Raw material supplier) to the ultimate procurer (consumer).

Now that we know what a supply chain is, let us understand the notion of supply chain management.

Definition of Supply Chain Management

Supply chain management is the art of aligning the supply chain with respect to a strategy and facilitate point to point connections between nodes of the supply chain optimally.

What is that strategy? Well, the strategy is the one that belongs to the strategic growth framework, more specifically, that part of the strategic growth framework which deals with the marketing framework. This explains why the supply chain management framework goes hand in hand with many of the strategies that fit in the strategic growth framework.

If we were to represent supply chain management mathematically, then we would define it as:

Supply Chain Management = f(Strategy, Optimized Logistics)

Optimized logistics is something that we call logistics that incorporates mathematical concepts to reduce waste. So what is logistics?

Definition of Logistics Management

Logistics management is the art of planning and delivering a strategy in real-time through facilitating entities befittingly.

The keywords ”delivering” and ”entities” are very very important. By delivering, we mean, projecting the strategy into reality, and by entities, we mean the participating nodes in our strategy.

Now that you have a good hold on what supply chain management is, we will now talk about the history and evolution of supply chain management.

The History of Supply Chain Management

The term supply chain management was first coined by British logistician Keith R Oliver in an interview in 1983. Although it was Keith who popularized this term, the concept dates back to 1950's, right after world war 2. There are seven stages with respect to the evolution of supply chain management:

Misconception - 1950s
Fragmentation - 1960s
Consolidation - 1980s
Integration - 2000s
Automation - 2010s

We shall discuss each of these stages in detail.

Misconception

This stage was outright after the world war 2. At this point of time, industrialists thought that the delivery stage of any product is just packaging and transportation. So you make a product, pack it and just transport it. Packing is relatively a ”noisy” concept for the context, i.e. packing protocols are pretty much covered in other areas, so we will focus more on the transportation aspect.

Transportation is no joke, we can't say that transportation simply means moving from one place to another. There is one factor that escalates the difficulty of transportation: ”Optimism”. Firms really need to optimize transportation costs, for which an enormous use of mathematical concepts is necessary. In this article, we solve the general transportation problem:

Assume that there are m different ways of transport, t₁, . . . , t_m, n different destinations d₁, . . . , d_n and distance D₁, . . . ,D_nand a total of T₁, . . . , T_n goods to be transported to the n destinations. Using transport t_i has cost c_i per mile and g_i number of goods need to be transported to destination d_i, while the total capacity that t_i can hold is C_i. Find the optimal way of executing this task.

The expressions that we want to minimize are:

∑

j=1

o_ijC_jD_i

where o_ij is the number of times transport means t_j has been used to reach destination d_i for all i. Another constraint we have for the above expression is that:

∑

j=1

o_ijC_j ≥ g_i

for all i because the transportation should be able to deliver all the goods successfully to the respective destinations.

To reduce lead time, the best option is to do all of it simultaneously, because definitely there will be enough trucks for transportation, and there is no use waiting for the truck to go and come and send them for delivery again, since the cost is the same, but the time is getting wasted. So the best solution for lead time reduction is to simultaneously deliver the goods to all dealers.

Fragmentation & Consolidation

In the 1960s, industrialists realized the need to segregate the delivery of a product into different dimensions, since purely transporting goods from one place to another would turn out to be a total waste if we did not take care of the more important aspects, such as inventory management. At that point, industrialists charted out the tasks that have to be carried out during delivering the product to the end consumers:

Demand Forecasting
Sourcing
Materials Handling
Inventory Management
Job Scheduling
Production Planning
Order Processing

The reason why this era was known as the fragmentation era was because each of these tasks discussed above were considered as separate tasks. Separate teams were set out for each task, and thus there was a lot of miscommunication between them. For example, the demand forecasting team has to report to the sourcing team who will report to the inventory management team and etc, and if some information is miscommunicated, it could lead to chaos. Moreover, teams would have no concrete idea on what they are supposed to do, as this separation between teams is ghastly and does not serve the sole purpose of integrated work. Now due to this, in the 1980s, industrialists realized that this separation is highly pointless and since there is too much tight coupling and cohesion between these tasks, industrialists merged all these tasks into one sole task: Optimized Logistics. Another important philosophy that was embraced by industrialists was that supply chain work inside the operations of the firm as well, so every process in the operations of a firm is part of the supply chain. Every internal team member with respect to the organization was considered a node in the network. This was one of the remarkable discerns that was made by industrialists and committees at that point.

At this point, all industries now turned their attention to the so-called Optimized Logistics Management. Below is a graphic view of the components involved in optimized logistics management:

As you can see in the figure, four of the seven components involved in optimized logistics are inside the Materials Management section and two of them are in the Physical Distribution section. With respect to a firm, materials management refers to cultivating and maintaining materials, while physical distribution refers to the distribution and flow part of the supply chain. The reason why Oliver made this distribution at that time was because he realized that to consider the entire supply chain holistically as one functional unit, it was necessary to categorize these fragments subtly into categories depending on the major functionality of the supply chain. Another important decision of his was to separate Inventory Management from both materials management and physical distribution. Since inventory management tries to fill the gap between them, it is partially materials management and partially physical distribution.

Now, to scent the beauty of this dissociation, we will first talk about materials management and physical distribution, and then connect these through inventory management. In each of materials management and physical distribution, we shall talk about why exactly these two major components contain explicitly, the minor components.

Materials Management

When it comes to materials management, as the definition states, this is related to manipulation of materials. If we want to manipulate materials, we need to know the supply that is required. What motivates the decision that finalizes our bill of materials? The demand of course. This is exactly why we need to forecast demand. Now, when you know the demand, you need to plan the requirements and the processes, which is why we have production planning included. What do you do after you have a plan? We need to source the materials of course, which is why we have sourcing. Finally, when the materials are sourced, we have to handle those materials, where handling is an activity that is considered to choose itself from transformation or transportation. Now that you know what exactly persists in the materials management framework, let's dig in deeper into the four components of this framework.

Forecast Demand

When we say ”forecast”, which literally means ”predicting the future”, we expect objects, heavily derived from the mathematical framework to be instantiated. What kinds of objects? To know about that, keep reading, and that will be covered in the upcoming sections.

Forecasting demand as a whole is a very complex concept. But to get a sense of how to forecast, we generally try to derive from the marketing 2.0 framework, which corresponded to Gen X, concepts that pertained to the creation of value. In short terms, forecasting demand was a process that undergoes a subtle ”give and take” policy.

Since most firms considered supply chain management as a separate functional unit, and not something that integrates with other departments, demand forecasting was not very successful, and planned attacks seldom existed.

So how exactly is demand being forecasted? To know that, keep following me to the next sections.

Production Planning

When we have some idea on the demand, and the final features of the product are finalized, we have to plan the requirements and the entire production process. This is a highly generic statement, when we enter the specifics for each industry, we find different techniques in each of them.

What we will do instead is try to write out some generic outlines as to how we can plan our production.

Firstly, in any production plan, there are three components:

People
Processes
Common Components

First, when it comes to people, there are some generic management techniques one needs to know. Firstly, and most importantly, the transfer of artifacts between people has to be minimized. We don't want someone to look up to someone else's work. Next, we need to keep track of people's work. Finally, as said by Mark Christopher, supply chain exist inside a firm as well, we must try to impart some basic supply chain management principle while transferring artifacts from person to person.

Second, processes are something that co-exist within each other and the entire firm. Now, we try to understand in quite a bit of detail, about process management.

Process management is the art of managing processes in ways that meet demand after successful forecasting of it. There are three types of processes:

Senior Processes
Core Processes
Supportive Processes

Senior processes are processes that govern and control the core processes and strategies in ways that keep the deviations in statistical business figures under control.

Core processes are the processes that actually convert inputs into valuable outputs. These processes can be considered as transformation processes.

Supportive processes are processes that assist the core processes, for example, parallel sub-processes in the chain of core processes.

The goal of the firm should be managing these three processes in ways that meet profitability and consequently demand. Senior processes should be very efficient and quick in reaction. Most senior processes abide decentralized management rules.

When designing core processes, there are three questions to be addressed.

What do we need?
Why do we need?
Who do we need?

The first question is in regard with the aspect that is lagging in the firm. For example, a car manufacturer lags windshields. The second question is in regard with the reason of that need. In the windshields situation, windshields are mandatory, because if it rains then drivers won't be able to see clearly. Finally, the third question is in regard with the set of ground workers who shall execute the task.

When these questions are addressed, we ask this set of questions again. Which means, in order to successfully attach the windshield in the car, the questions are asked once more, what do we need, why do we need and who do we need. These questions are asked step by step until the entire design is very clear.

Designing requires high level of organization, after each set of questions, the needs should be arranged in order, so when we reach the final needs, where no more questions can be asked, the entire process design is in order. Here is the example of the "what do we need" for a car windshield:

And so the root needs are reached and organized. Another important fact that we must consider is that, when we put '+' between two non-machinery attributes, it means there is another machine that assembles them together. So for the motor leaf, between wires and frame design, there is a second machine that puts them together. Finally, all the three nodes, wiper, motor and pivot are all assembled together to make a windshield.

This means, under a node, if there are n non-machinery attributes, there should be at least one more machine that combines those child-nodes together to make the parent node successful.

Drawing the design in the form of a flow chart always helps, because it gives a very good visual. Designing core processes are not the job of one man or one group. It requires the whole organization to work towards it.

I have also discussed the notion of supportive processes above, the process that is sub-designed under the core process that assembles more than one non-machinery attribute is called as a supportive process. Supportive processes can also be called as parallel dependent processes. One example of supportive processes is the HR training process that undergoes in any firm.

When we do this, we get a very good idea as to what exactly we want and how we are going to achieve it, and therefore, it is now time to move ahead to the next step which we must take: Sourcing.

Sourcing

Sourcing with respect to a firm is about sourcing external materials into the organization, i.e. purchasing materials. This may sound fairly easy, but as far as Optimized Logistics is concerned, it is not. Now, we present the sourcing problem.

Assume that you want to purchase M materials of the same type, and you have suppliers S₁, . . . , S_n, and suppler Si quotes cost ci per material. On order of x_i,j or more materials from supplier S_i, for 1 ≤ i ≤ n and 1 ≤ j ≤ f(i), where f is the function that sends supplier indices to the amount of schemes, you get a discount of d_i,j percentage. Each supplier S_i can supply at most T_i materials. What is the optimal way we can place our orders to these suppliers?

We want to optimize:

∑

i=1

Y_iC_i x 1_yi

where 1_yi is d_i,j , where j is the largest positive integer so that y_i > x_i,j . We also have the constraints:

∑

i=1

Y_i

as well as

y_i ≤ Ti∀ 1 ≤ i ≤ n

This problem is quite complicated, as we have indicator variables appearing in the picture. In the area of linear programming and regression, there is a technique that deals with indicator variables called sparse regression and convexifying non-linear programming problems. What we do here is basically use mathematical algorithms to simplify the use of indicator variables, which we will exclude from this paper, as it is already a huge research topic.

Once we source the materials, we move ahead to the next and the final component: Materials Handling.

Materials Handling

Materials handling can be considered as a "controlled" execution of production planning. The fact is that things may not go as expected, and that gap is what we call delta. To control that delta, we are calling materials handling as a controlled execution of our plan. Here is where an extension of process management comes into picture: Process Control.

The efficiency and effectivity of the process should be measurable, in which sense, there should be benchmarks which can point the level of process quality. The controlling of these parameters is called process control.

A vague picture of the measure in quality of processes can easily be figured out while proofreading the output. But the major challenge is to figure out the exact process and aspect that is getting out of control. In order to measure that, we have some techniques

Control Charts
5 Whys
Cause and Effect Diagram

Control charts are a pictorial representation of variations in the process. Control charts contain two limit lines, the UCL and LCL. If an attribute exceeds the control limits, then that attribute requires serious care, else we are good to go.

5 Whys analysis is a singular approach to to understanding the root cause of the problem by repeatedly asking 5 Whys. The answer you get in the end is the root cause.

Cause and Effect diagram is a multilingual approach to finding out all the root causes of an effect. Each and every aspect is discussed in detail in this analysis, unlike the 5 Whys which only goes to the root causes and does not understand the entire context behind it.

In contrast, process controls are a form of feedback and feedforward controllers, feedback controllers being taking action after the effect and feedforward controllers being taking action before the effect.

Investing on feedforward controllers is a lot more beneficial than investing on feedback controllers, because the effect doesn't even happen!

Physical Distribution

Physical distribution is focused towards the other half of the supply chain. There are two primary components in this framework, job scheduling and order processing.

I will elaborate more on what we mean by "Order". An order to a node is a situation that commands that node to execute a certain action, in this context, transmitting. So, an order does not necessarily have to be placed by a consumer, an order can be placed by any other node (including that node itself). Customer servicing is also an order.

Now, the first thing that happens in physical distribution is the wait for an order by any node. Once we get that order, we process that order and schedule distribution jobs amongst other nodes, for example, transport facilities. These are the two major components in the physical distribution framework, and now we shall dig deeper into them.

Order Processing

We must first note that the flow of information through the supply chain is as important as that of the materials.

When an order is placed, it is important to inform node that placed the order about the status of the delivery. Again, delivery here is a very broad term that tries to represent the expected response to an order. So, when a product arrived, it is a delivery, when a technician comes to service a repaired fridge, that is a delivery as well.

When an order has been placed, we need to first validate whether the order can be delivered or not. If not, we have to pass back information that the order cannot be completed.

Next, if the order can be processed, it is important to state the time taken for the order to be processed.

Once the order is processed, we have to inform the receiving node that the order has been processed.

These are the fundamental concepts of order processing.

Job Scheduling

Job scheduling in the context of optimized logistics is executing a medley of tasks in an optimal amount of time. Therefore, we present the generalized assignment problem:

Assume that we have to complete tasks t₁, . . . , tn and we have resources r₁, . . . , r_m. The cost to assign task t_i to resource r_j is c_i,j for all 1 ≤ i ≤ n and 1 ≤ j ≤ m. Find the most optimal way to execute these tasks.

We say that x_i,j = 1 if task t_i is done by resource r_j and 0 otherwise. Then, we want to optimize:

∑

i=1

∑

j=1

x_i,j . c_i,j

Which can, yet again be solved using linear programming techniques.

Inventory Management

As we said earlier, inventory management fills the gap between materials management and physical distribution, but why? First, let's understand the concept of an inventory.

The most common mistake that is made by people, is that people consider an inventory as a godown. This is what firms thought in the 1950s, i.e. the misconception era, but an inventory is a lot more than that.

Formally, an inventory is the element that connects two disjoint processes in the supply chain. In the entire supply chain, we have multiple disjoint chains, a product cannot be engineered in one go from the raw materials. When materials go through a process and transform into artifacts that concludes a chain, we need to store those artifacts in an inventory, and then those artifacts are pushed in as materials in another chain. The creation of artifacts falls under the materials management framework, while the distribution of artifacts from one chain to another falls under the physical distribution framework. This is exactly why we claimed that inventory management bridges the gap between the two frameworks.

In the optimized logistics framework, the primary objectives that we want to attain through inventory management are:

Easier locating of goods
Minimizing physical risk and motion waste
Coupling disjoint processes
Controlling inflow and outflow of goods

We are now ready to formally define inventory management:

Definition of InventoryManagement

Inventory management is the art of optimizing inventory costs, minimizing transportation cost and classifying inventory stock.

Minimizing transportation costs has been discussed in the earlier chapter, where we presented the generic solution to the generic transportation problem, so we will be talking more about optimizing inventory costs and classifying inventory stock.

Optimizing inventory costs

Inventory cost consists of three different costs:

Ordering cost - cost that is payed for ordering goods.
Holding cost - cost that is payed for storing the goods, such as, depreciation, deflation, taxes, spoilage, etc.
Shortage cost - cost that is payed when an order is placed by the customer but there is no stock in inventory.

The firm's key goal must be to minimize these costs. The first and foremost cost that must be minimized is shortage costs. The other two costs don't make direct impact of the firm in the market, while shortage costs may deduce customer retention rate. The second most important cost is the holding cost. Storing too much of inventory is a phase of waste, especially for expire-able products, since they have a deadline of usage. Most importantly, overloading of inventory causes blockages in the supply chain, which is dangerous for the firm. The final and the least important cost is the ordering cost.

Firstly, shortage cost should not exceed a penny. Secondly, holding cost must be minimized. That means, we can minimize both the shortage cost and holding cost by purchasing goods that are equal to the customer demands. This way, maximal goods shall take part in the supply chain, thus minimizing blockages. As professor showed us in the simulation exercise, minimal inventory is better than inventory overload. We can thus mathematically conclude that the reordering point would be when

Stock at point = Forecasted demand x Receiving time

and the reordering quantity would be

Reordering quantity = (Receiving time + gap) x Forecasted Demand + U

Here, receiving time is the time taken for the goods to enter the inventory. U here is the extra goods required in case there are any variations in the forecasting (buffer stock), and we are doing this because, shortage cost must equal zero. Finally, the gap is the number of days in which the reordering point appears. It is best that gap is minimized, because variations in demand shall cause blockages which can be painful. We have seen how industrialists worked with optimized logistics. Now in the next chapter, we will see how industrialists completely changed their thought process.

Classifying Inventory Stock

The most important differentiation between an inventory and a godown is that, an inventory classifies its stock and maintains a proper catalogue of it's items. How exactly does the inventory classify its stock? First of all, inventory stock must be classified based on the demand of products. There are different analytical techniques to classify stock, such as ABC Analysis and Movement Analysis which were discussed in my previous article, MBO in Investments. When we classify inventory stock appropriately, it becomes easier to locate items, and most importantly, keep the interfaces for delivering high-in demand products easier to access.

To conclude with, inventory management is one of the most important components in optimized logistics that minimizes blockages, or bullwhips in the supply chain.

Integration

Up till now, we saw how the world changed its views and perceptions on supply chain management from the second world war, till the 1980s, where industrialists considered supply chain management as a disparate functional unit. During this period, a double-dip recession hit the world, which caused severe inflation and unemployment. Why is it that this event occurred? What caused this recession? To understand that, let's first understand the situation that we can imply from the discussions above. As you might have noticed, firms at that point of time spend money on the delivery of a product without a foolproof plan. Why do we claim that they did not have a foolproof plan? Well that's simply because the forecasting of demand cannot be done without deriving concepts from marketing!

Now let us understand what causes a double-dip recession. But before that, what is a double-dip recession? Well, a double-dip recession is a recession followed by another fresh recession before the old recession occurred.

What causes the original recession then? Generally recession is caused when inflation increases, and inflation is caused when too many loans are pending to be fulfilled. The double-dip recession is caused when there is an imbalance between demand and supply, which would cause high levels of unemployment, or wages would be incredibly high.

Getting back to the story, at this point of time, most firms did not forecast demand, and to execute the optimized logistics, they took loans from banks. But in the end, firms went in losses, since the demand wasn't forecasted, and loans were pending to be fulfilled. When most firms are facing losses, there would be one firm where a lot of traffic hits, and that is where the bullwhip occurs in the supply chain, unlike other firms where demand was less and supply was high, here, it would be the opposite, demand is high, while supplies are less. This made the few firms that hit a lot of traffic to increase prices in products, due to which inflation occurred, and this combination caused the double-dip recession.

During this period in 1982, a man and his friend Tim Laser coined a term called Integrated Inventory Management, and the main idea behind that is to integrate optimized logistics with the other functional departments such as marketing and finance.

During an interview with electronics giant Philips, that man was triggered by a customer service manager Van T 'Hoff, where that man couldn't explain briefly about his idea of I2M. Due to this, he sub-consciously named his idea in that interview as... "Total supply chain Management". As you might have already guessed, that man is none other than Keith Oliver.

From that point onward, in the near 1990s, firms started to integrate supply chain management not only with the operations team, but also with teams such as marketing, finance and HR. This re-shaped the way firms forecasted demand, and now you shall see the second part of demand forecasting:

Forecasting is a concept that relies on three things:

Data
Mathematics
Approximations

Which can be mathematically represented as

Forecasting = f(Data, Mathematics, Approximations)

The most important aspect of all aspects in forecasting is the observational data that the firm has, in which sense, the data available about consumers on other firms. A combination of observational data and customer responses from appropriate questionnaires (which is carried out by the marketing team) shall reveal the problem that consumers are facing, and a vague figure of the demand that shall arise, for example high, medial and low.

To come out with more accurate figures, the firm must measure customer responses by the most challenging problem they are facing, for example, scrolling down to the bottom of a page on button phones. Another important fact that must be taken into consideration is that when a consumer invests his time into answering the firm's questionnaire, he has some genuine interest towards the firm.

Now the excitement and enthusiasm of the questionnaires responses shall be categorized into three:

High
Medial
Low

How do we measure excitement and enthusiasm? Well, excitement and enthusiasm are inversely proportional to the current comfort levels that the target market has with respect to the domain of our firm, so we can categorize the excitement and enthusiasm by reverse-engineering the responses given in those questionnaires.

Assume that h are highly enthusiastic responses, m are medially enthusiastic responses and l have a low level of enthusiasm. If the marketing and quality combination is done very well, then those h customers shall be the first to purchase the product at it's release. The next question that arises is, how many of the m + l customers shall purchase the product? In order to answer that, more questionnaires should be released to samples asking them about their expressions about products of other firms when those firms sent them a set of questionnaires regarding their product. Sample size shall be selected through concepts in sampling theory which shall not be discussed here. When a set of a customers react to medial and b of those purchased the product, then b/a of the medial customers bought the product, while in the low category, d/c bought the product. Circulating such questionnaires to samples about different firms shall yield the percentage of medial and low customers that actually bought the product. Taking the poisson distribution shall tell the firm the two percentage rates ɑ and β of medial and low customers respectively who bought the product after it's release. So, the demand estimate would be

ɑm

-----

100

βl

-----

100

and thus the demand is forecasted.

Note that this is a very simple model presented above, complicated models require gritty estimates which shall not be discussed in this article. From the observations above, we see how we instantiated objects from the marketing 2.0 framework, and how tight, marketing and supply chain management is.

This decision by firms is one of the most remarkable decisions, where optimized logistics are aligned to various strategies that are present in the strategic growth framework. This was the turning point in the evolution of supply chain management, and of course, the most historical age since the past fifty years of evolving, and evolutions to come.

Automation

While supply chain management did evolve well, in the early 2010s, we entered the digital era. At this point of time, firms really needed to move from the traditional way of supply chain management to incorporating technologies in supply chain management.

The first technology that was incorporated in supply chain was the IOIS, i.e. Interorganizational information systems. The IOIS tried to come close to the modern day technologies such as AI and IoT, but unfortunately, it failed, due to lower granularity levels. The IOIS had five levels of participation:

Remote I/O node - This creates point-to-point interactions between firms and a sensor on the local I/O enables actuators to begin the processing.
Application processing node - This takes into consideration the details given by consecutive firms for packaging and delivery, and executes the task as needed.
Multi-participant exchange node - This is a degree addition of the previous node. This node enables multi-application processing, from one point to different low-level points.
Network control node - This node automates maintenance of inventory levels, figures out potential errors in the network and fixes them, so it is basically a feedback controller.
Integrating network node - This node allows other firms to integrate themselves in the supply chain, yet keeps the entire processes above stable and secure.

The reason why it was less granular was because it couldn't manipulate information well. Now, we shall move on to the latest technologies, and how the escalate the quality of supply chain management. The two technologies that we will be talking about are:

AI & IoT
Blockchain

AI & IoT

The world is changing, earlier we used to have people doing physical jobs, such as delivery boys, drivers, cleaners, carpenters, technicians, and now we are going to have drones, self-driving cars, robots that clean entire houses and robots that can literally build themselves! The world ahead of us is going to be conquered by robots, robots shall replace humans, only innovators shall survive and the rest would be leading a miserable life! Drivers, cleaners, delivery boys, all go out on strikes, because if you don't, then you may not survive.

This is a common misconception that the entire world has on AI. Many even go to the extent of demoting AI & IoT and many went out for strikes. The fact that the common man is not realizing is that AI & IoT are just regular devices and have boundaries. As Grady Booch explained in his ted-talk, "To create a machine that can replace human, it would require way too much of training". So what exactly are AI & IoT, and why can't they replace humans even at the most mundane job ever? To understand that, let's first grind some fundamentals out of AI & IoT.

AI when expanded stands for "Artificial Intelligence". Before understanding what this term means, let us first understand what intelligence means.

Intelligence contains two components

Learning Ability
Analytical Ability

Learning ability is about the ability to process information, mostly inputs and outputs. As mam cited an example, you have cats and dogs. When you feed yourself those images, and the labels, you realise that "if it looks like this, then it is a cat, otherwise it is a dog". Throughout our whole life, we have been learning. How else does a kid speak his mother tongue without any training? The kid hears the input, and sees the output, and then realises what his parents are talking about. When his mom says: "Darwaja Khol", which stands for "Open the Door" in Hindi, the kid sees, that when the Hindi word is uttered, that particular action has to take place, so the next time his mom says the same thing, he will go and open the door.

Analytical ability is the ability to compare two different data sets and create an inference. When experienced HRs validate interviewees within seconds, it is not because they have magic inside of them, but they have seen multiple candidates. Here, they notice two different data sets, one, the set of answers to questions, and two, their performance. What they are doing here, is they are feeding themselves with data, which is then processed into experience, and next time when a candidate utters similar answers to a candidate that did not perform well, the HR wouldn't hire them at all.

Artificial intelligence is trying to push these abilities inside that of a machine. When we do so, this might allow machines to give valuable "predictions" using big data.

Now, what is IoT? Well, IoT, which is expanded as internet of things, is basically a data adaptor that converts a physical scenario, such as the tap of a button, into an expressible event inside the cloud.

Generally, when we talk about modern day technology, we cant seem to separate AI and IoT, and instead, we talk about the Interplay between AI & IoT. The IoT collects information and passes it on to the AI, which then takes a decision and the IoT expresses that information back, physically. For example, consider the case of looking for firearms. The IoT sensor that takes an X-ray of the luggage sends the X-Ray-ed image to the cloud, where the AI exists, and the AI validates the information. If there is a firearm, the AI reports back to the IoT, which then expresses the result: an alarm.

I will now answer the hot question: why can't AI& IoT replace humans. You see, the difference between us humans and machines is that, our brain is boundless. Humans have no boundaries, our brains can stretch to extreme limits, unlike machines, which have a boundary.

In every class of situations, there exists a layer of situations, which we call ”Noise”. These noisy situations are known as unexpected situations. When an AI encounters unexpected situations, its analytical results could create ruckus. How? To understand that, let's take the case of self-driving cars.

Assume that a driver is going across a street, and he encounters a man in the middle of the road, laying down, injured. Obviously, that man will stop the car and go help that man.

When it comes to self-driving cars, the IoT sensor of the self-driving car would send the information of something ahead in the road. The AI device might not have been trained to identify injured people that it would stop, and so instead, it considers the injured man as a speed breaker. The moment this decision is taken by the self-driving car, the car would go over the injured man. This is an example of why AI & IoT cannot replace humans.

Another famous example, is the worlds first ever robot-staffed hotel. The guests of this hotel had a nightmare of a stay, since the robots misinterpreted the snoring of a man as a call for room service and kept belling the guests all night.

The real question now is, how in the world then, do AI & IoT benefit us? The true benefit arises only when these technologies are used only where they are supposed to be used. If you give a carpenter a driller who uses it as a hammer, he would return with holes in walls, and of course, several smashed fingers.

This is where the human-machine pyramid comes in. The human-machine pyramid gives a clear distinction as to what the roles of a machine are, and what the roles of a human are. Here is the pyramid below. The blue sections are tasks that would be carried out by a human, and the green sections are those which should be carried out by machines.

Let us describe these six roles in detail:

Wisdom

Wisdom is something that is purely attainable only by a human, and wisdom is about making wise decisions. This stands on top of the pyramid, since this is the exact reason why we humans are much more intelligent than robots.

Insight

Insights are something that relate to the analytical abilities, the conclusions we can draw from situations. This is also something that has to be majorly taken care of by a human.

Knowledge

This is the repository of inputs and outputs that the machine stores, which it will use to take a decision in future.

Information

Information in this context is about converting the data that comes in different forms into something expressible.

Data

Data is just the set of inputs and outputs of a given situation in an unorganized way.

Noise

As described earlier, noise is an unexpected situation, which must be left for a human to handle.

As you may notice, we have made a clear distinction between the roles of a machine and the roles of a human. The reason why data, information and knowledge is given to the machine, is because machines are much better than humans at processing data. They can organize disorganized information way better than humans, and can easily deal with non-noisy situations. When we make use of machines from this angle, experienced is escalated. But how can we make use of AI precisely? Well, we can let AI handle very basic scenarios and use it as an advisor and not an actuator to slightly complicated scenarios which would then be validated by a human.

But how does all of this relate to supply chain? You see, AI & IoT primarily extend the original idea of IOIS, except that they provide better granularity. The usage of IoT in supply chain is primarily arched towards inventory management. IoT can help transfer data into a cloud which can then process and store data easily. We wouldn't need a supervisor that notes every single transaction, but instead, we would need a supervisor that "supervises" the information being processed. Using IoT, we can find the item that is the most moving, and when this information is fed into the cloud, the AI makes a recommendation to classify inventory stock differently. That recommendation is then validated by an actual human, and then implemented accordingly.

AI also helps in the forecast of demand. The marketing team does an additional exercise of sending questionnaires related to the products of other related firms, the AI then takes the data set of responses and the actual output (the results) that happened, and tries to create a correlation between those data sets. Once we do that, we feed the AI with the responses of the current market survey, through which the AI makes a prediction of the amount of demand. How does it make this prediction? We use tools from mathematics such as simple regression, polynomial regression, sparse regression and so on.

We now move on to concepts in Blockchain and how it is being used in supply chain.

Blockchain

Centralization is something that always co-exists, there is one guy that takes all the decisions. Blockchain aims to get rid of that central entity, the decisions are made by participating entities. The primary idea of a blockchain is to decentralize all centralized decisions which we, the public have no idea of. The blockchain idea was first implemented in currencies, the end product being bitcoin, which was published in

2008 by Satoshi Nakomoto. The main idea of bitcoin was to make secure transactions, transactions that were immutable, and validated by a consensus.

To express the concept of blockchain, we will align the main ideas of blockchain with bitcoin.

Firstly, there are three aspects when it comes to decentralized electronic peer-to-peer cash systems:

Transaction Validation
Immutability
Security

In addition to these, bitcoin provides us with an extra functionality which we call a smart contract, that would create a blueprint for the transaction to be executed. A smart contract plays an important role within blockchain. We shall understand how bitcoin attained these in detail.

Immutability

Immutability is something that is necessary in our day to day life, we need to make sure that information is legit. If the set of transactions that have been committed are tamper proof, then the whole ledger is trustworthy and note that each and every node has the latest copy of the ledger.

Satoshi presented a neat solution to execute immutability, that is, by introducing the notion of a block and the notion of a chain. A block is a combination of two objects, a header and a body. The body contains all the validated transactions under a time-limit, while the header contains metadata of the block, as in, it contains the hash value of itself, the hash value of the previous block, the time-stamp and the Merkle-root hash. Take everything granted for now, we will explain in detail how exactly they work, but just notice a property of the block: It contains the hash value of itself and the hash value of the previous block. How to find this hash value? That will be explained in the later sections, but as of now, consider hash value as an unique identification number. By containing the hash value of itself and the previous block, the two consecutive blocks get "chained". This means, all the transactions can be segregated into blocks with respect to time, and chained together. The entire ledger is distributed into blocks that are chained in a consecutive manner. The first block in this chain is known as the genesis block, which is a special block, since it has no previous block.

How does introducing blocks and chains help us here? Well, if an evil node wants to change any valid transaction, he must first change the block that contains it, but changing the block causes dramatic change in the hash value, thus breaking the chain! And that is a notification to all the nodes in the network that there is some problem, and the change in transaction shall be rejected by all.

Transaction Validation

Transaction validation is primarily about checking whether a transaction is a double spend or not. What is double spending? Double spending occurs when two amounts of the same value are transacted, but only one amount is deducted. This problem can be solved by introducing the notion of a timestamp server. If the latest digital signature and the timestamp (of when the transaction occurred) are the same for two coins, the miner shall reject the transaction. While if one miner accepts one transaction and another accepts the other, then by the longest chain algorithm, only one of those two blocks are going to be valid, thus rejecting the double-spend.

Since blockchain is decentralized, there needs to be some form of democracy in this transaction validation, how does that happen? This is exactly where the concept of consensus comes into picture.

There are two types of consensus, qualitative consensus and quantitative consensus.

Quantitative Consensus - Number of people deciding upon a single transaction's validity.
Qualitative Consensus - Systems deciding one person who decides a number of transaction's validity.

Qualitative consensus is better than quantitative consensus in many ways. One reason is scalability, but the major reason is, evil nodes can create thousands of alts, and the evil side will win the consensus, so quantitative is completely eradicated.

How does the qualitative consensus work? All the nodes in the network will be given a certain task, and the winner of the task gets to validate the transactions. What exactly is this task? Here is where the hash value of a block comes in. The task, that will be given to all the nodes is a computationally difficult problem, which is known as proof of work. The problem is to find a hash value less than or equal to the target hash. This can be done by taking the previous completed hash and adding a nonce to it and re-hashing the result. A nonce is an integer that is used only once. The difficulty here is to find the nonce value that gives us a valid hash that is at most a target hash. All the nodes work on finding the nonce value, and whomsoever finds the proper nonce value wins. What if two people found the nonce value in the same time? In that case, both of them get to validate the block, and there are two "brothers" in the chain.

This causes serious problems. To solve this, we adopt the ”longest chain” algorithm. What this does is, it takes the longest chain as the current state of the ledger, because nodes can choose to work on either one of the blocks, and the chain that's the shortest among both brothers would be simply broken.

How does all this guarantee that the ledger is valid? That's simply because, the plausibility of assuming that the winner of the proof of work is honest is equal to the plausibility of assuming that the individual who has a computer with maximal computing power is honest. And that's pretty true, because authorized entities only own quantum computers. The longest chain algorithm also shows that the longest chain between brothers is more valid, why else will honest nodes invest their time onto it?

Even though what we said above seems intuitively good enough, to convert it into code, it requires a lot more details. So, let us understand how hash functions work.

A hash function is a 1-way trapdoor function, which means, it's easy to go from the input to the output, but very difficult the other way around. Another important property of the hash function is that, any minor change in the input causes extreme changes in the output, in which sense, there is no pattern in the hash inputs and outputs. Another interesting fact of the hash functions is that, their output string length is precisely 256 bits, no matter how small or large the input is.

A question that arises is, since the length of the string is fixed, there are only finitely many possibilities, precisely 2₂₅₆, so what if, between some two inputs, a collision happens? That is,

H(X) = H(Y ),X ≠ Y

The best possible thing we can do, is to reduce the collisions. A hash function is more collision resistant when the algorithm behind it is very good. The hash function that is used in blockchain is called SHA256, that is, Secure Hashing Algorithm 256, and has been designed through the Merkle-Damgard construction. Let us understand this a little bit:

We begin by padding bits to a multiple of 256, which means, if the string length is a, then we pad -a (mod 256) more bits to the original string, and the additional b bits that we're concatenating, the first bit is 1 and all the others are zeroes. Assume that the resulting string has string length 256n, where n is a positive integer.

Divide the bits into n chunks, each containing 256 bits. Initialize a value, IV, which is 512 bit and choose a compressor function f, such that when two inputs are entered, and one of them is 256 bit, the output is 256 bits too. This can be done through logical, bitwise left and right operators. Apply f on IV and the first chunk, and you get a result. Apply f again on that result and the second chunk to get another result. Apply f yet again, on the final result and the third chunk and keep iterating this process until you reach the final chunk. Apply f on the final chunk and the result you got after doing this process n - 1 times to get the hash output FI. The figure above illustrates what we are saying.

Now, the question is, what is the basis of validation? We told that it's respect to time, but how exactly? First of all, the process of finding a valid hash for the block is known as mining, and the set of people who work on mining are known as miners. On average, a block is mined in 10 minutes, and within those 10 minutes, the transactions that have been committed would be validated, and the valid transactions

would be filled in the body of the block. The next 10 minutes would receive more transactions, those would be validated too and put in the block, and so on. An approximate of 2016 transactions are made within these ten minutes.

Security

Transactions between two parties must be uninterrupted, which means no third party can interfere. To ensure security between two parties, cryptographic algorithms, such as Diffie-Hellman, RSA and ECC are imparted.

The most common theme of all cryptography algorithms is the Public key, Private key algorithm. It's more of a ”lock and key” method, but let's follow standards for brevity.

The public key, private key algorithm works as follows:

Let A,B be the communicating parties and E be the eavesdropper.
A sends a lock to B, a copy of which is gone to E, since he's eavesdropping.
B uses the lock A sent and encrypts the message he wants to send, and sends it publicly.
A receives the locked message, uses his key and opens it, while E is busy trying to find out how he's going to unlock the locked message only using the lock.

The strength of the entire algorithm lies on how hard it is to find the key from the lock, and the crux of the algorithm is to be able to find the lock and the key.

Converting into mathematical terms, let's learn the RSA algorithm. As of now, we shall be talking in numbers, and not in words.

The RSA algorithm makes use of a very unique fact. That is, when p and q are sufficiently large primes, multiplying them maybe easy, but knowing that pq is the product of p and q is very difficult. Now, choose e such that gcd(e, φ(pq)) = 1 and find the multiplicative inverse d < ϕ(pq) of e modulo φ(pq). Here are the following steps for RSA:

A sends B the public key (e, pq).
B encrypts the message m as m^e (mod pq), the resulting answer that he gets is sent to A.
A encrypts the message by doing (m^e)^d ≡ m (mod pq), and thus the communication is successful.
Now, E has (e, pq) and m^e (mod pq). He won't be able to do anything with this, because to find m, he needs φ(pq), which cannot be found through pq, and thus the transaction is secure.

Although the RSA algorithm seems good, it takes a lot of CPU time. There's another algorithm known as ECC, which stands for elliptic curve cryptography.

Upcoming texts are going to involve pure mathematics, we begin with the notion of a field. A field is any set (possibly infinite) which is algebraically closed. The set of reals, rationals, complexes are a field, while the integers are not, because division is not closed. The field which we will be emphasizing more on is F_p, which is the set of integers modulo p, and remember, this is indeed a field.

We define elliptic curves over F_p as

y2 = x3 + ax + b

where 0 ≤ x, y ≤ p and 4a³ + 27b² ≠ 0. This expression is the discriminant of the curve equation above, and we don't want it to be 0, because we don't want double roots in our cubic.

Now to actually consider elliptic curves with the specified conditions, a field, it must be algebraically closed. So, we need addition, multiplication and inverses to be defined, and moreover the algebraic properties such as associativity must hold. Remember, in all the mathematical expressions, we take modulo p to make sure that we are talking in F_p.

We define point addition of two points A and B, as the reflection of the point C over the x-axis, which is the intersection of the elliptic curve and the line AB. How do we know if such a point exists? Well that's the property of the elliptic curve we chose! For any two points on the curve, there exists a third point on the curve such that those three points are collinear! Note that A + B = B + A here, because the line will anyway hit C. Proving associativity is a good exercise, as well as highly computational, hence we shall omit it in this text. For the reader's information, we have:

where all indices are taken modulo p.

Before we define point multiplication, let me define point doubling. Point doubling is when you add a point P to itself, which means, the line that we're considering is the tangent to P at the curve, and that intersects the curve at Q. Then the reflection of Q over the x-axis is the point that is doubled.

Let's suppose we want to define kP, where k is any positive integer, and P is a point on the curve. We take the binary representation of k:

where a is the unique positive integer such that 2^a < k ≤ 2^a+1 and ϵ_i are all integers in {0, 1} for all 0 ≤ i ≤ a. So, we have

We now calculate

where 2ⁱP is defined to be P being point-doubled i times. Then we use the associativity property of point addition to add the entire sum one by one and finally we reach our point Q. Again, indices are taken modulo p. Now why did we go for the binary notation? That's because multiplying a point by 2 is well defined, and associativity of point addition exists too, so by using binary notation we can successfully and uniquely calculate any scalar multiple of P.

All the heavy math is done, but how are we going to use all this into the public key private key algorithm? The experienced reader would have noticed that, if Q = k · P, and we know (P,Q), we won't be able to find k, which means going from (P, k) to Q is easy, but going from (P,Q) to k is very difficult. Although shor's algorithm in quantum computing has cracked this, not all hackers have quantum computers, so we're relatively safe. Whether this is super strong or not, it's surely a lot more secure and less costlier than RSA algorithm.

Let's now go to the elliptic curve cryptography algorithm:

A sends B the public key P = nG, where n is the private key and G is the base point on the curve.
B sends A the ciphertext, which is (kG,M + kP), where k is a private key known only by B.
A deciphers the ciphertext by doing
M + kP - n(kP) = M + knP - knP = M

and he successfully receives the message.
On the other hand, E has P and (kG,M + kP) and can't do anything with it, because to decipher this, he needs to know n.

This algorithm is a slight modification of the public key private key algorithm, since we have two private keys here.

We now have a concrete understanding towards encrypting/decrypting protocols. But cryptography in blockchain does not end there! Assume that A makes a transaction with B. To make sure that the transaction has not been interrupted, blockchain uses the digital signature algorithm, which is DSA. What this algorithm does is, using a public key which has been sent by the sender A, it will verify if the digital sigature in the transaction matches with the output of the algorithm. And the public key has been designed in such a way that when the transaction is interrupted, then the digital signature would have changed. And this implies that the public key has something to do with hash functions.

The most common digital signature algorithm, is the Elliptic Curve Digital Signature Algorithm, or in short, ECDSA.

The algorithm has three steps, generating the public key, creating the signature and finally verifying the signature. Let us understand them one by one:

Generating Keys. We set the elliptic curve for this algorithm to be a multiple of a sufficiently large prime p and let P = (x, y) be a point in the curve E.Generate a private key d ∈ [0, p - 1]. Compute Q = dP and let the public key be (E, P, p,Q).
Creating Signature. To sign a message t, we have the steps
- Choose k ∈ [0, p - 1] and compute (x1, y1) = kP, and set r ≡ x1 (mod p). If r = 0, restart the whole process.
- Compute
  s = k^-1(H(t) + dr) (mod p)
  
  If s = 0, restart the whole process.
- We now have our signature (r, s).
Verifying Signature. To verify the signature (r, s), we have the steps
- Verify that r, s ∈ [0, p - 1].
- Compute H(t'), where t' is the received message.
- Compute u = H(t')s-1 (mod p) and v = rs-1 (mod p).

Very well, how exactly do we implement this into blockchain? Well, the definition of an electronic coin is a set of chronological digital signatures. If an interferer tries to change the signature, then the algorithm will detect the change and thus reject the transaction.

The Merkle Root Hash

Computer science algorithms and graph databases play a prominent role in blockchain. When a node wants to validate any previous transaction in the ledger, it would be way easier to search the transaction, than continuously downloading latest copies of it, which is huge in size. Here is where Merkle root hash comes into picture. This is a graph database algorithm which allows us to efficiently sort transactions. Merkle root hash and B+ tree have a lot of similarities, the major similarity being the sorting principle. We would actually like to say that merkle root hash is the general hashed version of B+ tree, but we won't be discussing too much about the B+ tree, so let's dive into the algorithm right away.

For brevity purposes, assume that there are only eight transactions in a block, A,B,C,D,E, F, G,H. We have their hashes H(A),H(B),H(C),H(D). Then define

H(XY ) = H(H(X) + H(Y ))

and also H(ABCD) = H(H(AB) + H(CD)) and finally H(ABCDEFGH) = H(H(ABCD) + H(EFGH)). So the tree goes like this:

To reach to the destination of the transaction, we impart a similar idea as used in B+ tree, which is indexing. Here, the indexing is the concatenated transactions, so we can surf through indexes and easily land onto our desired transaction.

A question is, why do we need to keep all these complicated hashes? Merkle root hash has another job too, and it falls in the immutability category. When we said that if a transaction is interfered, the entire block is interfered, that is because when you change even one transaction, the merkle root of the block, which is the topmost vertex in the graph, i.e. the vertex from which the tree begun, changes. We can also figure out which transaction is invalid by checking the previous state of the block, comparing the hash values in the tree. One by one we descend down the tree by traversing through vertices whose hash values don't match and we find the transaction that has been changed.

Smart Contracts

Smart contracts are nothing but a piece of code which execute a particular task, provided pre-determined conditions are satisfied. In other words, they are contracts, but automated. Let's consider this example:

Alice and bob sign a smart contract that if bitcoin's value is more than ethereum's value in the next week, Alice shall pay 10 bitcoins, and Bob shall do the same otherwise.

Now, if the condition above is true, 10 bitcoins shall be automatically deducted from Alice's account. If it's false, the same shall happen with Bob.

This is just a vague example that we have taken, but you may wonder, what is the application of smart contracts in blockchain? What is the power of smart contracts? We would say that the power of smart contracts is huge, but for now let's understand how bitcoin smart contracts work.

When we, earlier told that an electronic coin is a chain of digital signatures, the signatures are signed on a smart contract which then deduct the amount from the payer's account. Which means, if A wants to send B 100 bitcoins, A signs a smart contract which then deducts 100 bitcoins from A and transfers it to B. Now to actually verify that the signature in the smart contract is the signature on the bitcoin, a miner uses the digital signature algorithm and validates the payment.

The applications of smart contract in bitcoin is not that vast, but it surely is when we change the point of focus to ethereum. Ethereum has a lot more features apart from those of bitcoin, some with more granularity, and some are completely new.

Some basic examples are running funds. So, five people have a common ethereum account and they have boundaries as to how much ether they can withdraw every day. Someone has one hundred percent rights, someone has five percent rights and etc.

A very good example is the internal consensus. What happens here is, there are a set of people (not everyone) in a common ethereum account and an internal consensus is held to take a decision, whether to keep a particular node in their private network or not. If majority says yes, then he is in, else he is out. Ethereum has also enabled liquid democracy delegation.

Apart from all this, we at gboxz family believe that smart contracts can be used to change the way this world works. Making violation of financial agreements practically impossible, and the concept of a smart contract can be given another dimension of extension, taking it from finance to other aspects of the world. But in this paper we will be talking only about the financial aspects.

Talking about finance, ethereum has incorporated financial derivatives through smart contracts.

What happens here is, when A has purchased the right, but not the obligation to buy some shares of an ethereum account (fund). Now, if A does not want to buy the shares, the smart contract stays silent, while if A does want to buy the shares, the amount needed to buy the shares is automatically deducted from A and transferred to B and A gets the shares of the company.

PBFT - A misconception

In many articles and websites across the internet, there are claims that show that PBFT is used for deciding a state of the ledger, but there is no notion of game theory in the bitcoin whitepaper. In the ethereum whitepaper, it is clearly written that "The main roadblock that all pre-Bitcoin currency protocols faced is the fact that, while there had been plenty of research on creating secure Byzantine-fault-tolerant multiparty consensus systems for many years, all of the protocols described were solving only half of the problem."

And it's very true, because PBFT assumes the fact that at least 75% of the nodes are honest, but this can easily be countered when all evil nodes create thousands of alts.

When the Byzantine generals problem was published, we were talking about individual people. One person cannot have two different faces in real life, so assumptions like above are plausible. But in the internet era, one person can have not two, but thousands of faces, making it practically impossible to satisfy the hypothesis.

Types of Blockchains

There are four kinds of blockchains:

Private Blockchains
Public Blockchains
Permission-ed Blockchains
Consortium Blockchains

Private blockchains are those blockchains that work within an institution. Public blockchains are blockchains where anyone can participate. Permission-ed blockchains are those blockchains where people that pass through certain regulators can participate, and finally, consortium blockchains are a collaboration of private blockchains, where the nodes that participate in the consensus are pre-defined.

Blockchain for supply chain

Now we come to the real question: how can we use blockchain for supply chain? Blockchain can be used in supply chain in a multitude of ways:

Getting rid of counterfeit markets
Imparting contracting between nodes in the supply chain network
Allows user to track their product

The first aspect is a very very sensitive one. As you can notice, many eatables and drugs are sold at a much higher price than the original one. So, when a counterfeit institution makes a transaction from a legit institution, the transaction is recorded, and therefore if the counterfeit institution increases the prices, by querying his transaction, we can figure out that the institution is counterfeit.

Contracts between nodes in the supply chain can be automated through smart contracts, as mentioned earlier.

Users can track their product from the manufacturer all the way to the post office, since every successful delivery is a transaction, and that allows users to feel more secure to transact with an organization.

Conclusion

supply chain management has been evolving ever since the second world war, and we got to see multiple drastic changes within the concept. The ways in which people looked at supply chain management, from an absolutely disparate fragmented functional silo, to a completely integrated robust system within all organizational activities such as marketing and finance. In the end, as all practices surrender to modern-day tech, the dominance of AI, IoT and Blockchain is apparent.

Technology for supply chain management is a holistic concept, AI& IoT help enhance precision in the day to day supply chain management decisions, while Blockchain enhances trust within an organization through immutability. Overall, supply chain management is an evolving concept, which shall continue to evolve, for time now, and for generations to come.

Quality Control

What is Quality?

Product quality, in intuitive terms, is a measure to how good a product is. But professionally speaking, we define product quality as

Definition of Product Quality

Product quality is a characteristic that defines the degree of the customer's exclamatory sense in a positive way.

Total Quality management as a whole can be divided into three different parts

Quality by Design
Quality Control
Quality Improvement

According to the father of quality, Dr. Juran, TQM, that is total quality management divides itself into eight different parts;

Customer focus
Customer focus
Process centered
Integrated systems
Strategic & Systematic approach
Fact based decision making
Communication
Continuous improvement

Let us briefly discuss each of these points:

Customer focus

Focus must be towards customer satisfaction. No matter how much efforts have been put into creating the product, if the paradigm is not customer-centered, then the product will not reach it's targets.

Employee involvement

Here is where MBO comes into picture, employees shall give their best investments only if they are motivated, have rights to make decisions, and note that the employees of a firm from all hierarchies represent the quality of the product/products.

Process centered

Lack of set processes of a firm are the major cause of fall in quality. The firm must have set processes integrated in it's skeleton, in order to increase quality.

Integrated system

Quality management systems play a crucial role in quality as they unite the design of a product with business processes in ways that benefit the firm and it's stakeholders.

Strategic & Systematic approach

In order for a firm to achieve it's business goals, there must be quality-centered strategic plans that help the firm improve quality.

Fact based decision making

The performance of a firm must be determined through business figures and immediate action-plans must be enforced to increase performance.

Communication

When a firm makes dramatic changes for business improvement, each and every employee must know the changes in order to invest their time and work appropriately that causes qualitative improvement.

Continuous improvement

Quality is all about improvement! Appropriate applications of quality principles installs kaizen slogans in the firm, thus causing drastic improvement in quality and increases business figures in a positive manner.

Earlier, we broke quality management into three different sectors, and now we have broken it into eight different crumbs. That means, each of these eight different crumbs belong to the three different sector in some sense. Thus, we classify the eight crumbs into the three sectors in the following way:

Quality by Design

Customer focus
Employee involvement

Quality Control

Process centered
Integrated systems

Quality Improvement

Continuous improvement
Communication
Strategic & Systematic approach
Fact based decision making

We are now ready to formally define TQM:

Definition of TQM

The art of managing the whole, to get a high degree of excellence.

What are the benefits that is yielded when TQM principles are applied? Here are the benefits:

Reduce risk
Prevent problems
Solve problems
Control processes
Reduce cost of quality
Increase productivity
Improve supplier performance

And the most important benefit of them all is to increase quality of the firm in all aspects, not just the quality of a product.

Although all this looks good, there is a cost that the firm must pay, which is known as the cost of quality. This cost can be visualized as:

The cost that is payed due to poor quality of a product is known as the cost of poor quality. An intuitive approach to minimizing cost of quality is to invest appropriately into prevention which reduces failure cost.

The quality department is now broken into three groups, G₁ who will work on quality by design, G₂ who will control quality and G₃ who will improve quality, and the common goal being to minimize cost of quality, and most importantly, cost of poor quality.

Quality Action Plan

Quality by Design

Quality by design is the art of designing a product that satisfies customers to the highest extent. Quality by design in quality management and creating value^* in marketing are very similar, except that creating value emphasizes more towards the communication side and quality management emphasizes more towards the delivering side. Quality by design can be called as part two of creating value.

What exactly is the function of quality by design in a firm? Quality by design strengthens three correlations:

Creator's perception about customer's needs
Expected product design and the actual design
Design and the final product

when graphed on a scatter plot.

Quality by design works more towards preventing failures, rather than detecting and fixing them, and as I said earlier, this reduces cost of quality and increases the customer base of the firm.

Quality of design, is broken into five tasks, which we wil call as the 5 D's. The 5 D's are:

Define
Discover
Design
Develop
Deliver

Let us understand each of these points in detail.

Define

The firm begins by defining the base foundation of it's product, selecting it's target market* and defining goals that are measurable.

The base-foundation for all industries are chosen by maximal area of expertise and innovative base ideas that differentiate the product amongst competitive products in the same area. For example, Zoho's CRM solutions and salesforce. Although the base ideas are the same, Zoho did differentiate between salesforce and itself by adding new features. When it comes to defining goals, both the service and manufacturing industries have a common list:

Business figures
Performance
Customer loyalty

What is customer loyalty*? In professional terms, it is the retention capacity of a firm.

Discover

The firm must now understand the needs of the customer, provided the base foundation of the product and the target market. Which means, the design of the product is a function of the customer needs and the base-foundation. Mathematically speaking,

Design = f(Base-foundation, Customer needs)

This makes the design of the product totally customer-centric. The design being customer centric alone is not sufficient, but the final product must be customer-centric too, which requires processes to be customer-centric. This ensures customer-centricity in the firm, thus embracing the first crumb of TQM.

A very interesting philosophy is the Pareto principle. This principle states that a small percentage of customers contribute to maximum revenue and the other large percentage contributes little. We break these groups of consumers as the ”vital few” and the ”trivial many”. In fact, this concept is not just applicable to customers, but also to root causes of problems, products that are liked by consumers and many more. Subsequently, we have the

Pareto diagram, that helps the firm visualize the vital few and the trivial many. This tool is pretty interesting, as it contains two graphs at once. The x-axis consists of attributes rather than numbers and the left side of the y-axis is the frequency, and the attributes are arranged in order, by the frequency count. While the right side is the cumulative percentage of total occurrences per attribute. Here is an example:

Design

The data that has been collected must be converted into product jargons*, which must then be converted into a product that satisfies maximum customers. This process of converting customer needs into a design must be better that those of competitors in order to outperform them. The firm must come up with creative ideas that please the customer. The question that arises is, how can the firm outperform other firms? For that, the firm holds analysis that reveals the features/ideas that other firms don't have and enforcing those features into the design. Some methods of analysis are:

Competitive assessment
Evaluation of alternatives

In order for the design to be perfect, totally customer-centric, multiple design-reviews must take place, the design must be looked at in a customer perspective, and this must be done by a set of people, not just one person, since the views may turn out to be biased. The design of the product must be scalable, which means the construction of the design must be taken from a general point of view, and not from a numerical point of view.

Development

A design is a visualization of the product, but to actually create the product, the firm must begin the development process. The process should be developed in such a way that:

Requirements are satisfied
Chronic waste is eliminated
Defect-free controls are incorporated

When it comes to the software industry, the only point the firm requires to be satisfied is the first one.

Delivery

This is the execution stage of the firm. We split this into two industries:

Manufacturing
In this stage, the firm must make sure that the processes are actually in production, and a scale-up strategy is implemented to increase production units.
Service
In this stage, the firm must have well-trained human resource that can serve the customers in a pleasing way. The hospitality of the service is like features of a manufactured product, so the major players in the service industry are the front end staff.

Quality Control

Quality control is the art of preventing problems and detecting problems that helps control processes in ways that satisfies customers, pertaining loyalty. Preventing problems is generally known as quality assurance. Quality control allows the firm to make sure that the product is in the right hands and helps the firm maintain changes and processes in a controlled setting, thus providing stability to the operational ability* of the firm.

Controlling the quality of a product is very important in order to understand consumer behaviour.

From a control theory standpoint, prevention is a feedforward controller, while cure is a feedback controller. Let us understand how the feedback controller works:

The total quality of each aspect in a process is recorded. Total quality includes factors such as completeness, time and etc.
The recorded data is then sent into a controller which also knows the quality goals of the firm.
The controller then measures the expected performance and actual performance of each aspect, and if there is a huge gap in between, it initiates an actuator (work of G₂) and positions the performance with the expected performance.
These steps are repeated again and again, which maintains a good control over the processes.

The overall feedback control process has two key steps:

Identify the problem.
Understanding the problem in detail, and thus solving the problem.

which we will be discussing in the coming texts.

To begin with, how does the firm measure the overall quality performance? Here is where control charts come into picture.

Control charts are charts that allow the firm to keep standard deviations in control provided the firm takes necessary actions to do so. Control charts have an upper control limit and a lower control limit, and a systematic and healthy process is generally considered when the data points are in between the control limits and get better and better when the data points go towards the average line. How do we calculate the upper and lower limits? Assume that we have a set of data points p₁, p₂, . . . , p_n over equally spaced time limits t₁, . . . , t_n. We find the standard deviation

where p is the average of the data points (y co-ordinates). We divide the area between the upper limit line and the average line (equation is y = p) into three different areas, each with width σ. Similarly for the lower limit line, we divide the region into three different areas with width σ. If, in the set of data points, a point pi lies on the first level, it's generally good. If it lies on the second level, it's not bad. If it lies on the third level, it's close to bad and if it exceeds the three levels, there is a problem. The goal here should be to bring maximal data points under the first two levels, and thus minimize future chances of data points exceeding the upper or lower control limit line.

Let's understand this better with an example. Assume that you are a team of six working for a project, and one week has elapsed. You want to find out if the work you are doing is productive or not, so you want to do a control chart analysis. The data points are

(1, 45),(2, 50),(3, 20),(4, 80),(5, 15),(6, 66),(7, 81),(8, 90),(9, 10),(10, 39),

(11, 50),(12, 90),(13, 56),(14, 87),(15, 68),(16, 87),(17, 88),(18, 34),(19, 45),(20, 66)

We begin by calculating the mean, which is 58.35. And then we calculate the standard deviation σ which turns out to be 25.36 and so our control limits:

I drew the LCL as 0 because negative LCL does not make sense.

The most important factor that we must take into consideration is that, the average working hours of the team in those twenty days must be close to the expected average working hours for performance. Because, if the average working hours is 55, and everyone works together for 10 hours, with very little standard deviation, then the control chart won't show any error. But that's an incorrect read, the working hours has been reduced by more than a quarter. So, we conclude that the average line must be around the expected hours for performance.

So the basis of deciding the excellence of a process is just two points

Gap between the average line and the expected performance.
Control chart readings.

Once the firm knows that there is a problem in a particular aspect such as overtime, production units and etc, by using control charts, the firm must understand the root cause of that problem (moving on to the second step of the feedback control process), because once the root cause is eradicated, there will not be much problems anymore. How can the firm find the root cause? There are some methods of analysis:

5 Why's
Cause and Effect Diagram

Let's understand each of these in detail:

5 Whys

The 5 Whys is a chain of ”why” questions linked with each other in order. Here's an example that professor discussed:

Albert failed the exam.
Why? - Because he wasn't feeling well.
Why? - Because he didn't sleep well.
Why? - Because he had a headache.
Why? - Because he rushed all the work in the end.
Why? - Because he didn't prepare in a planned manner.

And so we have reached to the root cause of the problem. Conducting a 5 Whys analysis seems trivial, but it actually is not, because going in a wrong direction of asking ”whys” shall lead to inconclusive discussions. So how can a firm conduct a 5 Whys analysis? Here are the steps:

Understand the problem

The most obvious step is to be crystal clear about what the problem is. For example, loose screws in products, subsequently causing defects. This must be seen by the analyst by his eyes so he is 100% clear about the problem.

Ask 5 Whys

The firm has identified the problem. The analyst must now ask 5 Whys to the ground workers and labourers. Let us take the loose screws example:

Products have loose screws.
Why? - Because the labourers are in a hurry.
Why? - Because there is a lot of work overload.
Why? - Because the manager is overloading them with work.
Why? - Because the senior manager is overloading the manager with work.
Why? - Because there is no set process for managing the work.

And thus the root cause of the problem is reached.

Find a solution

Once the root cause is found, the entire team must work for finding a solution to the problem, and discuss with the labourers directly their concerns and opinions on this. If the firm subsequently employs a feedforward control mechanism, then the poka yoke method, which doesnt allow the drill and the screw loose until the screw is tight shall be an excellent preventive measure. But this doesn't mean the 5 Whys analysis is useless here, through the 5 Whys analysis the firm understand that there is no set process for managing the work between the labourers. So, a combination of feedforward and feedback controls shall provide operational excellence in the firm.

Sustain

One time solutions don't work. There should be a team that monitors the operational excellence of that aspect, the example of discussion here being loose screws so that such errors don't occur in the future.

Cause and effect diagram

This is essentially the 5 Whys analysis but in a more in-depth and visual manner. Which means, the effect is divided into four major problems, and those four major problems are further broken into smaller problems, and the 5 Whys analysis is conducted into the small broken parts. So, the cause and effect diagram is a visual reduction of the effect, and thus helps all the teams related to the subject of the effect, discuss and come out with solutions and consistently monitor the processes, which shall eradicate the root causes, thus completely solving the problem. Before getting into solving the problem, the chains must be analysed to see if there are some logical mishaps, if there are none, the firm is good to go. Here's a visual:

And of course, when the root causes of the major problems are reached, an action plan must be designed to eradicate all those root causes, which will help the firm cherish, problem-free. More problems shall come later, but this problem identifying and root cause identifying analysis shall easily predict the problem, which means the firm must frequently conduct such an analysis.

Quality Improvement

Quality improvement is the art of increasing operational excellence. Before we understand what operational excellence is, let us understand what continuous improvement means.

Continuous improvement is the philosophy to continuously create and execute action plans in ways that benefit the firm. This includes:

Creating satisfactory products.
Increasing productivity. Reducing defects.

On the other hand, operational excellence is a broader form of continuous improvement. Operational excellence is achieved when each and every employee can identify the error in the processes and fix it. This requires high amounts of employee involvement, which means group discussions between all hierarchies must be frequent.

What are the characteristics that must be embedded in a firm in order to achieve operational excellence?

The 10 core principles of operational excellence are:

Treatment
All human parts of the firm must be treated with respect and without bias.
Employee Involvement
As discussed earlier, all employees must be given the right to contribute to the table.
Seek Perfection
As professor discussed, 99% is not enough, the firm must always aim for a 100% perfection.
Embrace scientific thinking
Scientific thinking involves
- Observe
- Ask
- Predict
- Test
Process Pessimism
The firm must always look at the worst aspect of the business processes. If any fault occurs, the processes must be rigorously reviewed and the issues must be identified.
Immediate Reaction
The first step of the firm is to ensure quality in the first place. If at all any error is found, then the reaction must be immediate.
Continuity of flow
The flow of the processes must be uninterrupted, in order to deliver high-quality products that satisfy the customer in real-time. If something is interrupting the flow, it is considered as a waste. For example, if one machine stops working, then that becomes a waste.
Systematic approach
As discussed earlier, there must be quality-centric plans in the firm in order to achieve business goals.
Sustained purpose
The purpose of the firm must remain the same from the beginning, and all the employees of the firm can work on the purpose foundation, and come out with new ideas.
Value for customer
Customer centricity, as discussed earlier, is the only reason customers shall retain. Therefore the firm must create values as the customers wish.

We have some methods of increasing operational excellence, which are:

Lean
Six sigma

Let us understand each of these concepts in detail.

Lean

The term lean, which means fat-free, or in a more general term, waste-free, is the process of eliminating waste through continuous improvement.

Let us understand this concept in three different areas:

Lean manufacturing
Lean services
Lean software development

Lean manufacturing

Let us first understand the different kinds of waste:

Overproduction
Waiting
Transport
Poor process
Inventory
Unnecessary Movement
Defects
Unused creativity

Before the firm begins the production process, the marketing activities of the firm should be able to predict the demand through mathematical tools such as regression and produce products up to the forecasted result.

The time taken in the process must be minimized, so the process must be built in such a way that the waiting time, that is the time taken for a machine or person to respond should tend to zero. This process can be constructed by having sufficient backup plans, such as, extra machines in case a machine stops working, or purchase high quality machines.

Transporting the product from one place to another is a time waste, because it adds no value to the product, in which sense, the time that is spent in transportation is a waste for the product, because the product, before transporting is the same as the product after transporting. This also applies to labourers and managers.

Processes are the heart of manufacturing. If the firm does not employ solid and systematic processes, the firm cannot scale, make changes or do any of the things discussed above.

Excessive inventory is not only a waste, but it can be dangerous too, when the products manufactured are expire-able, because they will waste money, more than time. Moreover, inventory causes blockages in the supply chain, so the firm must adopt the Japanese ”zero inventory” technique, which means, the products must be in motion, inventory is just like a guest-house for them. How can we find out the size of the inventory? This is discussed in the paper MBO in Investments, operational ability section.

Unnecessary movement for manpower can easily be eradicated if there are systematic processes that tell the employees and labourers the tasks they are supposed to complete by the end of the day. If the employees or labourers don't know what they're supposed to do, then they will do something that is not necessary, for example help someone else in shifting items from one place to another. Machinery movement can be eradicated if human movement is eradicated, and for the automated machinery movement, the instructions must be pre-set, in order for it to waste no time.

Defects can be reduced by taking preventive measures and proper inspection. As said earlier, a mix of feedback and feedforward controls allow the firm to reduce defects to the minimum.

Unused creativity is the idea or skillset which is not being used. It is considered as a waste, because if that idea or skillset is employed, then the product can benefit, so in product terms, it is considered a waste. This problem can easily be solved if there is proper employee involvement.

Looking back at the whole discussion we had just now, what actually matters for the customer here is high quality products delivered in minimal time. Another parameter that matters for the firm is limitation of resources. Let us first discuss the customer's expectations.

We define flow time of an unit as the average time taken for it to be produced. Subsequently, we define expected flow time of an unit as the time that is required to keep up the customer's demand. As a consequence, the expected flow time (in hours) can be given by :

Working hours per day

Demand in that day

The flow time and expected flow time must not have a huge gap, and so the question is, what must be the expected manpower? Let's pull some elementary math into this: Assume that one labourer can reduce the time taken to produce a product from indefinite to t. Then N such labourers can reduce the time to t/N. Let e denote the expected flow time.
Then, we must have

= e

soN =

In order to achieve operational excellence, the current state of the product must be measured. One tool for doing so is the current value stream map. This map documents everything going on in the process from the supplier to the customer.

This map contains widgets, each widget specifying the process and various factors that relate to the process. Those factors include:

Time taken for that process
Percentage defects for that process
Process description
Number of workers in that process
Delay time
Number of products produced within one shift
Issues in the process

And the overall performance of all the processes:

Cycle time
Value added time
Lead time
Demand
Expected flow time
Total percentage defect

Finally, the satisfaction of customers is also recorded.

Subsequently, to improve the current state map, a future state map is constructed, which automatically objectivizes the goals required to accomplish that future state map within some time limit, because the widgets are already divided, and each step of the process is also described, this allows the firm to understand the major reasons of the negative factors in processes, and doing a 5 Whys and a cause and effect analysis shall reveal the root causes, and a discussion for solving the root cause shall reveal the measures that are needed to be taken, in order to increase overall quality.

More than removing waste, lean is about organizing the business. A very interesting technique that businesses use is the 5S technique, which stand for

Sort - eliminate unnecessary tasks, equipment, etc.
Set in order - arrange the tasks in order, thus arranging the equipment based on the tasks.
Shine - keep the equipment clean, and moreover keep the processes required to successfully execute the tasks clean.
Standardize - create processes that organize the things discussed above.
Sustain - maintain those processes.

When it comes to lean servicing, the parameters of measurement change, for example, production here would be successfully servicing a customer, delay time would be customer waiting time and so on.

Lean software development

The major difference between lean manufacturing and lean software development is that, in manufacturing, the products must be repeatedly produced, that too physically. But in software development, one application is enough, the rest can be created into copies.

This does not imply lean software development is any easier than lean manufacturing, as there are other barricades:

Speed of functionalities
Scalability
Security

The code written for each functionality must be smart enough, so when executed, it takes very little time. To do so, there are various techniques in coding, some being rearranging pieces of code, some being introducing new variables which may seem unnecessary to the untrained eye, some being intelligently rewriting the piece of code. The programmer does all of this in order to reduce the time taken to execute the code.

After the product's release, the number of users using the product must be scalable, in which sense, while constructing the product, the parameters used must be generic, and not a fixed number. For example, if a product can hold a maximum of 100 users, then when 1000 users start using the product, the server shall be down, causing the users to not use the product anymore.

Network security is very important in a software product. Hackers can do anything, hack the application, which may cause great damage. To prevent network attacks, the product must embed cryptographic algorithms such as RSA, ECC and etc.

Before I classify waste in the software industry, let me explain how the software industry flow works:

The firm begins with the requirements of the product, which is designed from the customer responses and mathematical methods such as linear programming*. Let us assume that the requirements are r₁, . . . , r_n. Once these requirements are created, they are analyzed to decide if those requirements are feasible or not. Once the final set of requirements r₁, . . . , r_k are finalized, the skeleton of the product is created, and finally the programmers code the functionalities. Finally, after the product is made, it's tested, and compares the final product to the expected product. If there are any issues, the cycle repeats itself, starting from the analysis process. Subsequently, if the firm wants to update the software product, the process remains the same, except that it's easier than doing whole products.

Looking back at the software cycle, we can classify waste as:

Unclearity in requirements
Extra time for completing a task
Unnecessary code
Job-free employees
Unused creativity
Time delay
Poor processes
Lack of learning new concepts
Bugs

The very intriguing thing here is, suppliers are not even necessary. So, the value stream map consists mainly of the processes and the customer satisfaction/feedbacks.

The major difference between the software and the manufacturing industry is that, software development has some elements of uncertainty, while manufacturing guarantees certainty. This is because, visualizing a physical product is a lot easier than visualizing something abstract. For example, the guarantee that the application is scalable itself is pretty abstract. To ensure that the product will contain elements which can't be checked in a short-span of time, lots of testing and learning should be done. Therefore, if any decision is made about the product, the decision must be based on facts, and not abstract assumptions.

To speed up the process above, all the reading, learning, understanding must be done in the analysis stage of the process. This shall allow the firm to be able to consistently flow through the process, blocking any chances of ambiguity, thus reducing waste as much as possible.

Six sigma

The fundamental concept of six sigma is to place customer needs before anything, even the business. Six sigma follow DMAIC protocols for improvement. DMAIC stands for:

Define - be clear about the problem
Measure - understand the level of performance and customer satisfaction.
Analyze - find the root cause of the problem
Improve - discuss and select appropriate solutions
Control - check if the problem still exists and control accordingly

Applying six sigma to a process would result in just 3.4 defects over a million. Sigma stands for standard deviation, and six sigma reduces the deviation from little to none. This is because, in the control chart, 3σ is from the positive direction of the average line and −3σ from the negative direction. Six sigma is thus a proven method for improving quality standards of a firm.

Six sigma looks at quality from a totally different perspective, thus allowing firms to rethink about their quality strategies. In conclusion, deploying six sigma strategies shall result in

Optimizing equipment cost
Reducing time
Reducing non-preventive costs
More loyalty

Conclusion

Looking at quality from an amateurs eye may seem trivial, as in, quality may seem like creating a product that is very good. But little does one know the entire process for achieving quality. Quality, indeed is creating products that benefit both the firm and the customer, but if that definition is professionalized, it can be looked as ”fitness for purpose”, breaking this definition shall lead to areas that may not make sense unless connected in a systematic manner.

Summarizing the concepts discussed above, quality can be broken as follows:

Quality Design
- Define
- Discover
- Design
- Develop
- Deliver
Quality Control
- Control charts
- 5Whys
- Cause and Effect diagram
Quality Improvement
- Lean management
- Six sigma

Operations

What is Operations Management?

Operations are the heart of a firm. Operations basically mean the mandatory tasks that a firm must commit to, in order to provide salaries, satisfy shareholders and pay off debt. For example, operations in a manufacturing firm would be production. Knowing operations, we are now ready to formally define operations management

Definition of Operations Management

Operations management is the art of managing operations through transformation processes in ways that keep the firm stable.

I would like to clarify two points here, firstly, stability of a firm stands for the ability of a firm to reach its business goals. Secondly, transformation processes are process that convert an input to a valuable output. The term valuable here is very important, because converting inputs into garbage is not at all useful for the firm.

In order for the firm to be able to measure, and thus improve the operations, there should be control over the quality of inputs, transformation processes and outputs. The vague solution goes by introducing feedback controllers in every step, so the operations flowchart goes like this:

Operations management basically manages this flowchart. We can split operations management into two different sectors:

Manufacturing
Service

The operations in manufacturing can be broken into product design and inventory management while the operations in service can be broken into operations control and scheduling.

Overall, both industries contain process management, supply chain management, resource planning and demand forecasting in their operations management exercise.

This is visually drawn in a flow chart below

Let us understand each of these points:

Manufacturing

Product Design - converting the requirements, needs and specifications into a manual for labourers.
Inventory Management - keeping the inventory levels as low as possible, which minimizes blockages in the supply chain.

Service

Service Design - converting the requirements, needs and specifications into a manual for customer-servers to follow.
Waiting Management - managing the customer service allocation for all customers, which means, reducing wait time.

Common

Process Management - syncing business goals and processes in ways that keep the customers satisfied.
Resource Planning - minimizing usage of resources and allocating them in the most beneficial manner.
Supply Chain Management - managing the flow of goods or services from the origin to the customer. Flows in services may include contracts and sub-contracts.
Forecasting - ability to visualize business figures in future. Variations may occur, resisting those variations come under process management.

Between any two processes or steps in a process, there are feedback controllers such as quality controls that keep the variations between the business goals and the actual results in this entire flowchart to the minimal. In order for the controls to be executed perfectly, continuous analysis must take place. To learn more about feedback controls, refer to MBO in Product Quality.

Detailed Action Plan

Product & Service Design

In order to come out with a design that satisfies maximal customers, the firm must first understand the needs of the customers. For that purpose, we have various methods of collecting data:

Observational Research
Ethnographic Research
Questionnaires
Qualitative Research

more details about these methods are discussed in MBO in Marketing. Back to the point, using such methods of analysis, the firm can extract the customer's needs, which also involves converting responses and results derived from the analysis into product jargons and then using modelling concepts such as linear programming that satisfies maximal customers. This concept is very well known as creating value in marketing management.

While working on the product design, the firm must take some aspects into consideration:

Reliability
Robustness
Degree of originality

Let us understand each of these points:

Reliability

Reliability is the measure of how well the product or service fits for the prescribed purpose. This is basically a part of product quality, and reliability can be improved by

Testing
User education
Multiple design reviews

The reliability of the entire product is a function of the reliability of each functionality or component in the product and the degree of user education, so in mathematical terms:

R(Product) = f(R(f1), . . . , R(fk), Degree of User-Education)

where R(t) denotes the reliability of component/functionality/product/service t.

In order to improve reliability of each functionality, the firm must conduct in-depth analysis and measure the degree of excellence in every aspect of the product while designing the product and the execution process. To learn more about this, visit MBO in Product Quality.

Robustness

Robust need not necessarily mean the strength of a physical item. Robustness of a product is the ability of the product to be able to function in multiple states/conditions. For example, consider two balms, one that only works for cough & cold and the other one works for severe headaches, sprains, twists, fever, etc. The second balm is definitely more robust than the first one, since it can be used in multiple states/conditions.

In the name of robustness if the needs of customers is sacrificed, then there is no point in keeping the product robust. Robustness must act like an extra layer of customer satisfaction, for example, customers would prefer a fragile phone with high features than a non-fragile phone with lesser features.

A good example of products that continue keeping customers satisfied while being robust is Tesla's truck. The windows are so strong that it would require three hammer hits to successfully break it, but the aesthetics and features of the truck are well maintained.

Originality

If the firm is looking forward to improve an already existing product, for example Samsung after iPhone, then the firm must conduct a multiple alternative analysis to the pre-existing product to find the extra features that can be added which shall cause diversion in customers. Pricing plays a very big role too, for example if Samsung priced it's phones high, it wouldn't have gained the market share it has today.

Originality of the design can be tested by the percentage functionalities that are not existing in the market up to date. Again, completely original products with low quality don't gain any market share in the product, instead someone else copies the ideas, increases quality and wins the market.

Inventory Management

To begin with, inventory is a place where manufactured goods ready to be sold are stored. Note that there is a huge difference between inventories from the trade sector and the inventories from the manufacturing sector. Inventory in the manufacturing sector can alternatively be defined as the conjoining element between a sequence of processes that is broken into two disjoint processes. So, when there is a gap between a sequence of processes, that is the sequence is broken into two disjoint chains, the element that fills the gap is an inventory. While the inventory in the trade sector is a warehouse where the products from the distributor are stocked.

Inventory in general is not a dump where products are stored. Inventory is a systematic arrangement of goods that help the firm in

Locating goods
Minimizing physical risk and motion waste
Coupling disjoint processes
Controlling inflow and outflow of goods

We are now ready to formally define inventory management:

Definition of Inventory Management

Inventory management is the art of optimizing inventory costs, minimizing transportation cost and classifying inventory stock.

Let us understand each of the three points in the definition above in detail.

Optimizing inventory costs

Inventory cost consists of three different costs:

Ordering cost - cost that is payed for ordering goods.
Holding cost - cost that is payed for storing the goods, such as, depreciation, deflation, taxes, spoilage, etc.
Shortage cost - cost that is payed when an order is placed by the customer but there is no stock in inventory.

Stock at point = Forecasted demand × Receiving time

and the reordering quantity would be

Reordering quantity = (Receiving time + gap) × Forecasted Demand + U

Minimizing transportation cost

Cost here includes both money and lead time. Lead time is the time taken for goods to travel between two consecutive elements in the supply chain. What we are trying to do here is to reduce transportation waste. To do so, we have a mathematical model, which is known as the transportation model. Let us convert the general transportation problem for monetary cost into mathematical terms:

Assume that there are m different ways of transport, t₁, . . . , t_m, n different destinations d₁, . . . , d_n and distance D₁, . . . , D_n and a total of T₁, . . . , T_n goods to be transported to the n destinations. Using transport t_i has cost c_i per mile and g_i number of goods need to be transported to destination di , while the total capacity that t_i can hold is C_i . The expressions that we want to minimize are:

∑

j=1

o_ijC_jD_i

where o_ij is the number of times transport means t_j has been used to reach destination d_i for all i. Another constraint we have for the above expression is that:

∑

j=1

o_ijC_j ≥ g_i

for all i because the transportation should be able to deliver all the goods successfully to the respective destinations.

Classifying inventory stock

Organizing inventory is one of the most important aspects when it comes to inventory management. Because if everything is messed up, then it would be very difficult to identify the goods needed to be transported and it would take hours to locate the product. Following the 5S principles shall allow the firm to organize the inventory. The question that arises is, how does the firm know, which products are most valuable and which are the least? For that we have two methods of analysis:

ABC Analysis
Movement Analysis

To learn more about these, visit MBO in Investments.

Let us now look at the visualization of inventory management:

Waiting Management

Waiting lines occur when there is an imbalance between supply and demand in timely terms. If the servicemen are waiting for job, then there is less demand, more supply, whereas if customers are waiting for their service, it is more demand less supply.

The costs that the firm will incur when there is an imbalance in demand and supply are:

Waiting space cost
Customer dissatisfaction
Salary waste
Loss in customers

When it comes to services, the supply of manpower that service customers is dependent on

Forecasted customer demand
Working hours
Difficulty of tasks

Working hours

Manpower = f(Demand,Working Hours, Difficulty)

Basically, the capacity of the firm is inversely related to the customer waiting time. Similarly, the demand of the customers is inversely related to salary waste. I am using the term related and not proportional because proportional follows equivalent ratios, while relations necessarily don't. Before I determine what exactly (or at least approximately) f is, let us understand some concepts.

First, we understand the characteristics of waiting lines:

Channelization
Queuing Order

Channelization is basically the process of distributing customers arriving for one purpose into different channels, while queuing order is the discipline that is followed for servicing customers. A very common and intuitive discipline is the ”first come, first serve” basis.

We now define customer arrival and service rates:

Arrival Rate - the rate in which customers arrive in every unit of timely measurement per channel.
Service Rate - the rate in which customers are serviced in every unit of measurement per channel.

These two rates shall be denoted by µ and λ respectively. Before going any further, here are some results:

Customers Served - The average customers that are being served is

µ

λ
Customers Served per channel - The average customers that are being served per channel is

µ

λ × Number of Channels

Waiting Time

Assume that the timely measurement is t hours and µ is the arrival rate in the timely measurement t. That means, on average, a new customer comes in every t/µ hours. If

< s_t

where s_t is the servicing time per channel (in hours), then the waiting time is zero. So, assume that

≥ s_t

Then, the amount of time the n^th customer waits is

W_(t,n) =

- s_t

hours per channel. If the number of channels is C, then the wait time reduces to

W_(t,n) =

- s_t

Expected Probability of Number

The expected probability that there are n customers in the waiting line in average per channel is

P(n customers in waiting line) =

That's because,

P(n customers in waiting line) = P(n − 1 customers in waiting line)

and the expected probability that no one is in the waiting line is given by

1 -

Expected Customers in line

We know that the expected probability of n customers being in line is

So, if the expected number of customers in line is E_c, then,

E_c

= 1

so Ec is a solution to the discrete logarithm above. We can better rewrite it as:

E_c = log_λ/µ

Before I approximate the function f, we understand that for each service si in the firm, the total manpower is C, the number of channels.

We are now ready to explicitly define f:

Let µ_i denote the arrival rate per channel for service s_i and λ_i be that of service rate with timely measurement t. Then,

where t(.) is the time function in hours and w is the working hours. The 60/13 came because each customer cannot wait for more than thirteen minutes, which means that the final customer in each channel should get a maximum of thirteen minutes in waiting time. Therefore, we set n = E_c and W_(t,Ec) = 13. The last step follows because services are generally open for ten hours.

An interesting fact is that,

λ = f(µ, channels)

which derives itself trivially from the equations above.

A very natural question that arises is, how can we calculate λ_i and µ_i? Before I proceed, let me introduce you to the notion of poisson distribution.

Poisson distribution determines the probability of a situation occurring x times which suits our problem of calculating λ_i and µ_i. In order to effectively evaluate these values, some amount of data is required to perform this act.

Consider a sample containing discrete data. If the curve of that sample is a function f, then the poisson distribution of that curve is the point x on the curve such that

where ρ = E[x] = Var(x), which means that ρ is the expected value of x such that it equals it's variance.

Applying this on the samples of arrival and service rates in timely measurement t shall give us the rate in which customers arrive in timely measurement t and the rate in which they are serviced.

Forecasting

Forecasting in the context of economics is the process of cultivating supply appropriately that keeps the supply chain healthy. Which means, there are no blockages, or shortages.

Forecasting is a concept that relies on three things:

Data
Mathematics
Approximations

Which can be mathematically represented as

Forecasting = f(Data, Mathematics, Approximations)

Taking back a look at the waiting management section, this actually makes sense, because to evaluate the λ_i and µ_i, we needed data, of course there was immense mathematical touches and most importantly, we were not super accurate with the model, and we can't be, because of the unpredictability in the behaviour of arrival rates. Although service rates can be controlled, because it's based on the firm, arrival surely is a lot less controllable than the service rate.

Forecasting can be successful if and only if, the customer inputs are converted into an output, which means, the consumer behaviour and responses should be converted into a superior product with high quality measures. Converting customer inputs into an output is basically the crux of marketing management, so this section will enforce a lot of marketing concepts.

The most important aspect of all aspects in forecasting is the observational data that the firm has, in which sense, the data available about consumers on other firms. A combination of observational data and customer responses from appropriate questionnaires shall reveal the problem that consumers are facing, and a vague figure of the demand that shall arise, for example high, medial and low.

Now the excitement and enthusiasm of the questionnaires responses shall be categorized into three:

High
Medial
Low

Assume that h are highly enthusiastic responses, m are medially enthusiastic responses and l have a low level of enthusiasm. If the marketing and quality combination is done very well, then those h customers shall be the first to purchase the product at it's release. The next question that arises is, how many of the m + l customers shall purchase the product? In order to answer that, more questionnaires should be released to samples asking them about their expressions about products of other firms when those firms sent them a set of questionnaires regarding their product. Sample size shall be selected through concepts in sampling theory which shall not be discussed here. When a set of a customers react to medial and b of those purchased the product, then b/a of the medial customers bought the product, while in the low category, d/c bought the product. Circulating such questionnaires to samples about different firms shall yield the percentage of medial and low customers that actually bought the product. Taking the poisson distribution shall tell the firm the two percentage rates α and β of medial and low customers respectively who bought the product after it's release. So, the demand estimate would be

h +

αm

100

and thus the demand is forecasted.

Note that this is a very simple model presented above, complicated models require gritty estimates which shall not be discussed in this article.

Process Management

Process management is the art of managing processes in ways that meet demand after successful forecasting of it. There are three types of processes:

Senior Processes
Core Processes
Supportive Processes

Senior processes are processes that govern and control the core processes and strategies in ways that keep the deviations in statistical business figures under control.

Core processes are the processes that actually convert inputs into valuable outputs. These processes can be considered as transformation processes.

Supportive processes are processes that assist the core processes, for example, parallel subprocesses in the chain of core processes

Core processes should pertain a customer-centric design, which can be done by enforcing quality principles. There are three aspects when it comes to core processes:

Process Design
Process Improvement
Process Control

Let us understand these aspects in detail:

Process Design

This is the first step of the core process. Before designing a process, the firm must address a few questions:

What do we need?
Why do we need?
Who do we need?

And so the root needs are reached and organized. Another important fact that we must consider is that, when I put '+' between two non-machinery attributes, it means there is another machine that assembles them together. So for the motor leaf, between wires and frame design, there is a second machine that puts them together. Finally, all the three nodes, wiper, motor and pivot are all assembled together to make a windshield.

This means, under a node, if there are n non-machinery attributes, there should be at least one more machine that combines those child-nodes together to make the parent node successful.

I have also discussed the notion of supportive processes above, the process that is subdesigned under the core process that assembles more than one non-machinery attribute is called as a supportive process. Supportive processes can also be called as parallel dependent processes.

Process Control

The efficiency and effectivity of the process should be measurable, in which sense, there should be benchmarks which can point the level of process quality.

Control Charts
5 Whys
Cause and Effect Diagram

5 Whys analysis is a singular approach to to understanding the root cause of the problem by repeatedly asking 5 Whys. The answer you get in the end is the root cause.

Investing on feedforward controllers is a lot more beneficial than investing on feedback controllers, because the effect doesnt even happen!

Process Improvement

Process improvement is equivalent to quality improvement. I am adding the texts of MBO in Product Quality on the Quality Improvement section here.

Continuous improvement is the philosophy to continuously create and execute action plans in ways that benefit the firm. This includes:

Creating satisfactory products.
Increasing productivity. Reducing defects.

What are the characteristics that must be embedded in a firm in order to achieve operational excellence?

The 10 core principles of operational excellence are:

Treatment
All human parts of the firm must be treated with respect and without bias.
Employee Involvement
As discussed earlier, all employees must be given the right to contribute to the table.
Seek Perfection
As professor discussed, 99% is not enough, the firm must always aim for a 100% perfection.
Embrace scientific thinking
Scientific thinking involves
- Observe
- Ask
- Predict
- Test
Process Pessimism
The firm must always look at the worst aspect of the business processes. If any fault occurs, the processes must be rigorously reviewed and the issues must be identified.
Immediate Reaction
The first step of the firm is to ensure quality in the first place. If at all any error is found, then the reaction must be immediate.
Continuity of flow
The flow of the processes must be uninterrupted, in order to deliver high-quality products that satisfy the customer in real-time. If something is interrupting the flow, it is considered as a waste. For example, if one machine stops working, then that becomes a waste.
Systematic approach
As discussed earlier, there must be quality-centric plans in the firm in order to achieve business goals.
Sustained purpose
The purpose of the firm must remain the same from the beginning, and all the employees of the firm can work on the purpose foundation, and come out with new ideas.
Value for customer
Customer centricity, as discussed earlier, is the only reason customers shall retain. Therefore the firm must create values as the customers wish.

We have some methods of increasing operational excellence, which are:

Lean
Six sigma

Lean

The term lean, which means fat-free, or in a more general term, waste-free, is the process of eliminating waste through continuous improvement.

Let us understand this concept in three different areas:

Lean manufacturing
Lean services
Lean software development

Lean manufacturing

Let us first understand the different kinds of waste:

Overproduction
Waiting
Transport
Poor process
Inventory
Unnecessary Movement
Defects
Unused creativity

Processes are the heart of manufacturing. If the firm does not employ solid and systematic processes, the firm cannot scale, make changes or do any of the things discussed above.

Defects can be reduced by taking preventive measures and proper inspection. As said earlier, a mix of feedback and feedforward controls allow the firm to reduce defects to the minimum.

We define flow time of an unit as the average time taken for it to be produced. Subsequently, we define expected flow time of an unit as the time that is required to keep up with the customer's demand. As a consequence, the expected flow time (in hours) can be given by :

Working hours per day

Demand in that day

The flow time and expected flow time must not have a huge gap, and so the question is, what must be the expected manpower? Let's pull some elementary math into this: Assume that one labourer can reduce the time taken to produce a product from indefinite to t. Then N such labourers can reduce the time to t/N . Let e denote the expected flow time. Then, we must have

So, N=

In order to achieve operational excellence, the current state of the product must be measured. One tool for doing so is the current value stream map, which visualizes the entire process. Subsequently, to improve the current state map, a future state map is constructed, which automatically objectivizes the goals required to accomplish that future state map within some time limit, because the widgets are already divided, and each step of the process is also described, this allows the firm to understand the major reasons of the negative factors in processes, and doing a 5 Whys and a cause and effect analysis shall reveal the root causes, and a discussion for solving the root cause shall reveal the measures that are needed to be taken, in order to increase overall quality.

More than removing waste, lean is about organizing the business. A very interesting technique that businesses use is the 5S technique, which stand for

Sort - eliminate unnecessary tasks, equipment, etc.
Set in order - arrange the tasks in order, thus arranging the equipment based on the tasks.
Shine - keep the equipment clean, and moreover keep the processes required to successfully execute the tasks clean.
Standardize - create processes that organize the things discussed above.
Sustain - maintain those processes.

When it comes to lean servicing, the parameters of measurement change, for example, production here would be successfully servicing a customer, delay time would be customer waiting time and so on.

Lean software development

This does not imply lean software development is any easier than lean manufacturing, as there are other barricades:

Speed of functionalities
Scalability
Security

Before I classify waste in the software industry, let me explain how the software industry flow works:

Looking back at the software cycle, we can classify waste as:

Unclearity in requirements
Extra time for completing a task
Unnecessary code
Job-free employees
Unused creativity
Time delay
Poor processes
Lack of learning new concepts
Bugs

The very intriguing thing here is, suppliers are not even necessary. So, the value stream map consists mainly of the processes and the customer satisfaction/feedbacks. Six sigma. The fundamental concept of six sigma is to place customer needs before anything, even the business. Six sigma follow DMAIC protocols for improvement. DMAIC stands for:

Define - be clear about the problem
Measure - understand the level of performance and customer satisfaction.
Analyze - find the root cause of the problem
Improve - discuss and select appropriate solutions
Control - check if the problem still exists and control accordingly

Six sigma looks at quality from a totally different perspective, thus allowing firms to rethink about their quality strategies. In conclusion, deploying six sigma strategies shall result in

Optimizing equipment cost
Reducing time
Reducing non-preventive costs
More loyalty

A lot of the concepts discussed here are highlighted. To learn these concepts in detail, please visit MBO in Product Quality.

A problem that firms face is process variation. Process variation occurs when the design of the process, which is dependent on the forecasted demand varies from the actual requirements and customer demand. Variation can occur in four different sources:

Forecasting
Assignable
Variety
Buffer

Forecasting variation occurs when the forecasted demand and the actual demand are pretty far. This can cause either blockages or shortages. In order to prevent forecasting variation, observational and ethnographic research must take place.

Assignable variation is when the inputs are defective, causing the process control to shake. Preventing this variation requires a feedback controller which verifies the quality of inputs and rejects inputs if they are defective. Of course, this requires a machine as well.

Variety variation occurs when there are too many requirements in products and services. This variation can be eradicated if the 5S principles are incorporated in organizing the tasks and requirements.

Buffer variation is the variation that cannot be completely eradicated. In order to minimize this buffer variation, the entire burden goes towards the processes themselves, so quality improvement techniques must be implied towards processes.

Resource Planning

Resource planning is the art of identifying, organizing and minimizing resources in ways that satisfy the needs of the firm and the forecasted demand.

Identifying resource needs is known as material requirements planning or MRP 1. This concept has already been discussed above in the process design section, where we consider the product structure tree and go to the root needs of the product. Organizing resource needs has also been discussed in the product structure tree, where we align the needs of each requirement in order and have a machine that assembles two or more non-machinery resource needs.

What is left to be discussed in resource planning is minimizing the resource needed. For this purpose, we have some models which work in optimizing cost. The first model is the assignment model. Assume that there are r resources a₁, . . . , a_r and n tasks b₁, . . . , b_n and the cost of using resource a_i in task b_j is c_ij for all 1 ≤ i ≤ r and 1 ≤ j ≤ n.

What we want to minimize is the expression:

∑

1≤i≤r, 1≤j≤n

c_ijε_ij

Where ε_ij = 1 if resource a_i is used in task b_j and 0 otherwise.

Note that this assignment model requires the tasks to be flexible, in which sense, any of the resources can be used in those tasks. For example, creating a wiper and creating a motor for the windshield are two inflexible tasks while creating a wiper and a stylish wiper are flexible, because what matters is the frame and the design can be coded into a computer program.

In the name of optimizing resources, quality should not be sacrificed. In a set of flexible tasks, the resources offered should be over a buffer quality limit. Now the question is, what if a resource is slightly high in cost of a resource which is much lower in quality but is at least the expected quality limit? Of course, the resource which is much higher in quality and slightly higher in cost is a lot more worth than the other resource. In this case, we are only losing. The assignment model fails in this case. The solution to this problem is to create a measurement function f, which when inputted any particular resource gives an output as a numerical measurement. Coming out with such a function requires huge knowledge in mathematical modelling, as there are various parameters to be measured which is definitely, rocket science and hence is omitted in this article. Assume that we have such a measurement function f which gives us quality measures of the resource. In that case, the expression we will be minimizing is

∑

1≤i≤r, 1≤j≤n

c_ijε_ij

f(ai)

Supply Chain Management

When a manufacturing firm comes out with a super-high quality product, the firm expects equivalent profits for their work. But coming out with a high quality product and staying idle never works. In fact, totally managing the firm itself wouldn't lead to much profitability. What matters here is if the super-high quality product has landed on the hands of the customer properly or not. This requires managing the entire supply chain point-to-point, including the firm itself.

As I said earlier, supply chains are chains that visualize products moving from one firm to another until it lands on the hands of the customer, we also looked at how to eradicate blockages and shortages, which keep the supply chain healthy.

But supply chain management isnt just knowing what goes where, when and how. Supply chain management is the process of managing every aspect in firms in the supply chain, point-to-point.

The internal customers of the manufacturing firms are the point-to-point firms which belong to the supply chain.

There are three components when it comes to supply chain management:

Conformance
Relationships
Inventory Management

Conformance is the process of guaranteeing that the required products and number are reached to the desired firm on a point-to-point basis.

Supply chain relationships are the relationships between point-to-point firms in the supply chain, which when is strengthened can cause profitability to all the firms in the chain.

Inventory management, as discussed earlier, is the art of optimizing inventory levels for all the firms. This is surely a difficult task, because the chain must flow from the first point (manufacturing firm) to the last point without blockages or shortages. Looking properly at this, one can realize that with proper time management, inventory levels can be minimized for all the firms.

This looks pretty intuitive, because in a chain, there are three components:

Widgets - Inventory Management
Connections - Supply Chain Relationships
Flow - Conformance

Conformance can be guaranteed through models like the transportation model in addition to some machinery which looks out for defective products at it's arrival.

Supply chain relationships should be strengthened, in which sense, there should be no delay in arrival of goods, often discussions between point-to-point firms and etc. This keeps the internal customers satisfied, and their satisfaction is reflected on the customer's face.

In the information era, supply chain management has incorporated a concept called IOIS. IOIS, which stands for Interorganizational Information System is a system that uses information technology to overcome physical difficulty in informing. Basically, IOIS is a software technology that automates conformance and inventory management between point-to-point firms in the chain. IOIS consists of five levels of participation:

Remote I/O node - This creates point-to-point interactions between firms and a sensor on the local I/O enables actuators to begin the processing.
Application processing node - This takes into consideration the details given by consecutive firms for packaging and delivery, and executes the task as needed.
Multi-participant exchange node - This is a degree addition of the previous node. This node enables multi application processing, from one point to different low-level points.
Network control node - This node automates maintenance of inventory levels, figures out potential errors in the network and fixes them, so it is basically a feedback controller.
Integrating network node - This node allows other firms to integrate themselves in the supply chain, yet keeps the entire processes above stable and secure.

Conclusion

Operations management has been one of those concepts where knowledge comes from experience and not from literature. A lot of topics (which aren't included here) cannot be explained in simple words, they require real-time experiences. While what can be explained in words is the top layer of operations management and a overview of each of the concepts in detailed. There are six different aspects when it comes to operations management, four of which are common and two of which are specific. So, we have six groups of people working together in each sector, call the Gi for all 1 ≤ i ≤ 6.

The overall objectives thus are:

Product Design or Service Design

Reliability
Robustness
Originality

Originality

Optimizing inventory costs
Minimizing transportation costs
Classifying inventory stock

Waiting Management

Reduce salary waste
Reduce wait time

Forecasting

Response measuring
Observational & Ethnographical research

Process Management

Senior processes
Core & Supportive processes
- Process design
- Process control
- Process improvement

Resource Planning

Optimize resources
Materials resource planning

Supply Chain Management

Conformance
Relationships
Inventory Management
Using IOIS