In Chapter three I had lightly touched on the aspect of biases when Planners try to side with certain decisions as opposed to others, this predilection for certain outcomes is not without any reason and it pervades every organizational interface and influences results. We shall see in the limited space of supply chains how such biases impede decision making in a direction that may or may not be the ideal one and taking cognizance of this early enough is an important way that some mitigation efforts could be attempted.

While psychologists have theorized on Bias theories, they have looked at individual and social biases much more deeply while organizational bias is a subject still in its infancy and I would take the liberty of using some of my experiences in explaining these biases.

Why do various constituencies hold certain perspectives in preference to other equally valid perspectives? First of all we must therefore define under what conditions such preference is desired and the first condition is:

1.       When there is one to choose from many while no obvious choice is available

This is a very general condition of choices, but it is the fundamental condition. Unless we have a situation when an obvious choice is available the presence of bias does not creep in, however there could be a social bias or an individual bias or a cultural bias towards certain choices although an obvious choice is available. We shall ignore this bias as this could be more difficult to handle in the context of organizational puzzles that we want to deal with. So we shall simplify the structure as one in which no mathematically true statement is available to make a particular choice obvious, we have the incidence of a ‘Bias’ to creep in.

Let us explore a given condition that a firm has three choices to increase profits and it needs to select one:

–          Increase Sales

–          Increase Margins

–          Reduce Costs

Let us assume that each could equally be effective to result in a higher profit, what would the firm choose? Perhaps in this case there could be an incidence of a bias towards either sales or margins or costs depending on what is predominant in the firm’s collective thinking. Most firms try to eliminate this bias by focusing on more than one objective function by designing certain incentives that serve various aspects of the original objective function.

2.       When functions are aligned to common incentives, they mirror specific biases instead of reducing it

Let us go back to the earlier example and assume that the firm puts in place incentives to ensure that all three objective functions are pursued, which means that increasing sales, reducing costs and increasing margins are all taken up by various arms of the firm so that we do not miss the opportunity to improve on each of these potential gains possible to be derived for increasing profits.

In order to avoid that the sales function single mindedly pursues the objective function of ‘increasing sales’, while the production function would pursue the reduction of costs, and the common group of people drawn from sales and plant to pursue the common objective function of increasing margins as that could be done by producing more of the high margin products or move towards increasing prices of some products by delivering higher value to the customers, we could have one single incentive directed towards profit which is to be achieved by the optimization of all three objective functions.

Objective Function Sales = S

Objective Function Cost = C

Objective Function Margin = M

Conventional logic would suggest that under clear methods of calculation we could actually pinpoint how each delta change could impact the final result which is profit, thus:

Profit = f (delta S, delta C, delta M)

Does this mean that the common incentive on profit could eliminate the bias towards either of S, C or M?

If this was true then the sales function would have had no bias towards sales, the manufacturing function would not have had any bias towards cost and the common function would have had no bias towards any one of the three at all. The truth is somewhere in between, we do see such biases although common incentives try to eliminate such biases. Why is this so?

The reason is steeped in the concept of information asymmetry and in what we have already seen a glimpse in the earlier chapter, Bounded rationality under uncertainty. Let us deal with each of the concepts first.

Kahnemann and Tversky researched on cognitive biases and have proposed three kinds of heuristics in play for decision making under uncertainty:

  1. Representative Heuristic Bias: Suppose we start with a set of data that represents Type A and let us assume that some characteristics of this data is related with a function B. For example let us take the typical example that Kahnemann and Tversky proposed; there is a list containing names of prominent Statesmen (Type A data). There are some within this who are more prominent than the others (Function B: Popularity). If the Type A list is read out then it is most likely that the more popular names would be remembered and the less forgotten, therefore it is more likely that attributes corresponding to the more popular ones would influence the decision on what is the representative attribute for the Type 1 data. For example if more popular Statesmen are men in that list, it could well be interpreted that the list of Type A has more men, regardless of the actual composition. Or a different example could be that if the names of popular Statesmen happen to be from Asia, it could be interpreted that the list has more proportion of Asian Statesmen.
  2. Availability Heuristic Bias: In this case attribution is given to a more familiar and resonant piece of information in the puzzle that is easy to decipher, than the more complex one and is more likely to be sequestered. The classical example is the bat and the ball puzzle, which was given to the Harvard University MBA students in which more than sixty percent gave the wrong answer:

If a bat and ball together costs $1.20 and if the bat is costlier than the ball by $1, what is the cost of the bat?

The availability heuristic has the biggest connection in the world of business. Investors find in the trend charts of the past performance of various companies a leading indicator of financial health of a sector of business or the economy in general, whereas the actual data could be simply random. The availability heuristics are more commonly used by the financial community to look at the data of the nearby quarters to frame the projection for the next quarters, which in absence of any other information is the only way to project future performance. The problem is that there are a myriad of other extraneous factors that could have influenced the more current quarters and whether they could influence the future needs to be ascertained before such an analogy could be drawn.

There is a different connotation to this problem. Investors do not have the full picture in a presentation, what they could have in a balance sheet or a detailed earnings statement with supporting charts. What a presentation conveys is a view point and how this gets presented creates the availability heuristic as it is never possible for the investor to discern a barrage of information steeped in data tables. Thus the guidance given in a presentation in case it is biased towards a set of trends, it helps to create a stronger impression for a particular decision to be made. Human minds never want to make complex calculations to get to decisions and the normal bias is towards known and tested ways of taking decisions which is based on simple heuristic judgments.

In the world of planning availability heuristic plays havoc with decision making as problems are so complex that planners do not have the full picture in absence of data and takes into account those that are simple to comprehend. Algorithms are the only logical way to look at more than one conflicting objective function and create proper computational logic to deal with optimizing of each against the other or the overall.

  1. Anchoring and Insufficient adjustment: In this case the focus shifts towards a particular initial starting point and then works backwards or forwards around that starting point to arrive at the best estimate. So the initial starting point influences the decisions considerably as the mind cannot make a sufficient adjustment in absence of new information. If we ask what year Man landed on the moon, one would probably start with the tenure of Kennedy to start with or his assassination date, if known, as there is a strong association of Kennedy with the launch of the first manned aircraft.

In the field of business this has wide ramification as in most business presentations one would see this bias rampant. The starting point of period under review or the comparator used, or the choice of metric itself influences the decisions, which is based on the anchoring that happens in our minds. For example if a period under review is shortened with the starting point as one year back versus five years back, changes the perspective completely. A large delta Vs a small delta could be simply because of the starting point and influences the anchoring and the subsequent insufficient adjustment.

Let me now try to use these concepts in the problem that I referred in the context of a common incentive approach with three objective functions playing to the fore. But there is an additional bias to be dealt with, which is the bias of the ‘mindset’.

While the general concept of mindset is about paradigms that give groups of people their comfort zones to deal with problems, to which they have familiarity in terms of solutions, there is the typical case of the Einstellung effect, which creates the ‘functional fixedness to problems’. If a salesman has spent his time more with customers in the field, he would not see the internal perspective of an operations function to deal with a problem that could have little to do with the market and vice versa.

If profit comes from the summation of efforts in sales, operations and strategy/planning, but if the specific components could not be dissected to de-construct the profit function, it is a fertile ground for cognitive biases to take over. On top of this there could be other objective functions like growth that could further vitiate the dynamics. One would have expected detailed mathematics to take care of such problems but managers do not have time with complex Mathematics, in fact they detest them. Rather they work with simplistic structures that simplify the problem and uses aids to facilitate decision making by trial and errors.

The three functional groups in question clearly have to deal with their mindsets and also the organization reporting to them. What typically happens in a common incentive structure is that the sales function maximizes sales, the operations function tries to minimize cost and the strategy / planning cell given the directive on growth looks at investments that could create growth, which may or may not be accretive to the bottom line in absence of good quality of data.

Information asymmetry in this context is a more interesting phenomenon. If we go back to the used car problem first referenced by Akerlof, we see that the buyer has no idea whether the car is a lemon (defective) while the price is fixed leaving no room for negotiation. If we assume that there are 20% lemons it could well be that the used car salesman sets the price in such a way that he maximizes his utility by the superior knowledge that he carries related to the quality of the cars against the inferior knowledge that the buyer carries. The solution to this problem is signaling.

There are many stunning examples of information asymmetry problems like managers versus investors in the knowledge of the earnings projections and the quality of these projections or the more extreme case of the cost of capital itself as the bid-ask spreads could expand due to one side not having the real picture while the other side could have.

The more blatant market example is the case of institutional investors having far better information than the normal individual marginal investor and could significantly gain from information asymmetries (short selling has been found to be an institutional phenomenon which has disproportionate gains and losses).

Let me go back to my example of the specific biases in the context of the common incentive program. There are actually many possibilities, so let me draw the decision matrix as follows:


One would notice that this is fairly simple in a two by two but in actual organizational context this is a multiple of X by Y, where quite a number of objective functions have to compete with each other.

It is a little difficult to exactly frame the entire matrix when there are more than two to be looked at a time. But how does the organization dwell on the common fringe areas where one conflicts with the other? Is the mere common incentive for one common objective a good enough mechanism to take care of the problem? Does leadership have to play the role of intervening to set the right priorities, but based on what information and data would the leadership act?

In most cases complexity in data structures actually forces biases to act and in this case all the three biases, including the ‘mindset’ acts and actually impedes the process of achieving the common objective as conflicts are resolved not based on what the problem requires as a solution, but by one or more biases that act more powerfully over the other.

3.  When the Owners of the Process come from specific functions and expertise, the process- changes reflect predilection of the owner

Think of a department being run by an engineer or a science graduate as opposed to an accountant.

Recently in a conclave in Zurich, this example was cited as no other than Azim Premji, the Chairman of Wipro, and he said the first thing that an accountant would do if he sees some numbers on a piece of paper against certain attributes, he will add them up; the engineer would not.

I have tried to think how mental predilections work in a given situation. It is different from bias, although there could be some common elements in both. Imagine a sales function being executed by an Engineer, he would invariably bring in a technical aspect in selling, try to understand the customer’s process from the technical process stand point, or communicate with his internal operations organization with a technical linkage in order to improve the process flow with the customer. This same function would look different when steered by a salesman who is coming from an accounting background or a purely selling background.

When such owners combine in a matrix set up to deal with challenges that stretch beyond their areas, it is only to be expected that predilections do play a role in both framing, understanding and the final solution to the problem.

Let me now go back to the original Supply Chain problem with multi-player situation. Is there a way to deal with the ‘push-pull’ that one would expect related to bias and predilections? And how do we deal with that?

The Organization

The organization is the center piece of the puzzle and what is the organization? It is just not the structure, but the people who man this structure and the relative importance of what they deliver or their functions deliver against each other and against the objectives of the firm. While designing a supply chain solution, the question of the organization can never be left out.

Every part of the organization has its core objectives and goals and in today’s world of incentives, they have assumed far more importance as the variable pay component has increased. In firms where incentives are so designed to maximize each individual function’s goals or targets, the problem is only compounded to that extent as overall optimization can only be done at a higher level. But how high is the higher level that is the critical question. The lower the better as better coordination can only happen if objectives can be commonly denominated.

A supply chain has many players starting with:

  1. Procurement of inputs, who have to negotiate against tough conditions and have their own KPIs like reducing transaction cost, procurement cost or improving cash flows through better payment terms. In a global organization such players have to be globally coordinated as they have to deal with global players as rivals as well.
  2. Sales: In a world where demand is constrained they behave quite differently to a place which is supply constrained. The sales function has to deal with a myriad of factors which I will cover in the next section.
  3. Supply Chain Planner: The supply chain planner works under the narrow influence of his planning horizon and have to depend on many other constituencies on whom they have little influence in terms of driving decisions, like production, maintenance, etc, who may have other KPIs.
  4. Production & Maintenance function: The operations function outside of planning actually influences results most prominently once a plan has been frozen as they take decisions outside of the planner’s view point and sometimes these decisions have bearing on the results.
  5. General Management: The General management is not an extrinsic player but the most intrinsic and he could take the ultimate decisions which may not touch base with the planner or its knowledge. While these decisions are important and could be absolutely rationale given the nature of the KPIs, some of these decisions could alter the way the results are supposed to be derived.

There are a number of processes that have to be followed through to make a planning process work. Let me highlight the key ones:

  1. Business Planning and Annual Budgeting Exercise
  2. Determination of  Dynamic sales and Production Plan
  3. Capacity Allocation and Supplier Quota Allocation
  4. Monthly Forecasting process
  5. Order Entry and Order Management
  6. Supply Demand matching
  7. Conversion of Flexi orders to Fixed orders
  8. Execution of plan
  9. Mid-course correction and adjustment of master parameters
  10. Expediting and deviation management
  11. Negative feedback loop

How the organization works in all these planning domains is an important step to be seen and debated. Do all parts of the organization have a common objective while working in each planning step? How do we take guard that any misalignment could be removed at every step? Can this be done simply through KPI monitoring?

When deviations to plan happen not because of random causes but assignable causes how do the system and all the constituencies come to know of it? How does the information flow work? Can this be done through ‘Rule adherence’ matrix only?

These are some key questions to be asked.

The Selling Mindset

Why I have singled out the selling function as a separate section, needs some clarification. This is where the ‘people who make stuff’ start the process of value creation in any firm. Without an order nothing can start, without a perceived demand, no other idea can have any meaning. While it the significant source of value creation, it is also a source of cost creation as much of the costs starts to accrue once the ‘package of delivery’ is conceived, or negotiated and finally agreed with the customer. No matter what we try to do in terms of designing a supply solution, the starting of the contractual agreement with the customer is the real starting point of how the customer would expect the supply to look like. The expectations we imbibe in the customer must take into account the inherent nature of the supply chain capability that we have, which is easier said than done, as the selling function has to deal with a large number of other factors that come to the fore. These are:

– Competition and competitive offerings

– Nature of the market

– Nature of the business

– Timing of the boom-bust cycles

– Objectives of the firm vis-à-vis growth and strategy in specific segments

– Power of the customer

The nature of offerings and the capability to deliver is a subject of debate within the firm; it is never concluded as selling part of the organization sides with the customer in many aspects of the debate and the rest of the organization must succumb to the realization that the customer must have its way. There is nothing wrong in this belief, other than the fact there are limits to capability and its costs and more often than note the offering is a step above what the costs can bear and this fact is never in black and white, while the offerings continue to be.

The planning mindset on the other hand must deal with optimization puzzles that either maximizes or minimizes a given objective function; there are no absolute truths to be pursued, as there aren’t any.

The operations mindset to maximize efficiency come in sharp conflict and therefore the need for overall leadership is of paramount importance.

In the context of Planning therefore I would ask, that the organization must try to improve its processes, but it should not start with a promise that has embedded in it semblance of a wishful thinking that can only be fulfilled through a series of initiatives. Rather it should start with what the organization is capable of doing and then start to improve on it through carefully orchestrated steps over a period of time. Sometimes this aspect is ignored while designing a supply chain process and we mix up between preparing to supply the ‘best of our promise’ with an inadequate capability to organize our inputs or our capability to produce. To find a decent balance to this puzzle needs organizational collaboration and removal of biases on which the chapter is dedicated.


Procyon Mukherjee


1st March 2011

Cognitive Biases And Their Impact On Dealing With Conflicting Objective Functions
Tagged on:

Leave a Reply

Be the First to Comment!

Notify of