Cause-and-effect as the ABC of practical logic

Outlining clearly the causality behind undesired-effects, and wondering what effects, desired or not, would be caused by the actions we take, have been an integral part of TOC from its start in the early 80s. In the early 90s several structured procedures were developed by Dr. Eli Goldratt, in the format of cause-and-effect trees, called the Thinking Processes.  I think it is time to experience the merits, but also the limitations, of using logical claims in the shape of ‘Effect A’ is causing ‘Effect B’, for managing organizations.

My bachelor degree was in Mathematics, which is the ultimate use of strict logic.  In our daily practice we use logic both to reveal the causes behind effects we experience and also for speculating what is going to happen if we take a certain action.  However, that use of logic is not easy; it is combined with a lot of emotions that confuse the strict logic.  Even when we do our best to stay within the logical directives we are faced with several obstacles.  One of them is being able to distinguish between assumptions about cause and effect and actual causality. We certainly have great difficulty with hidden assumptions, meaning not being fully aware that the causality is only assumed and not necessarily valid.

Reality is fuzzy and includes huge number of variables that have some impact.  In order to live in such reality we have to simplify the picture we have in our mind.  We do it by ignoring many variables, assuming their impact is too small to truly matter.  The choice of what we ignore is part of the basic assumptions behind our cause-and-effect logic.

To experience the value and the boundaries of applying cause and effect let’s check the following effort to understand a practical logical argument.

It seems straight-forward logic to claim:

If ‘We improve the availability of items on the shelf from 80% to 98%’ then ‘Sales will go up’.

Is this assertion always true?  Are there some missing conditions (insufficiencies) for the causality to be true?  Even if it is true can we deduce how much more sales will be generated?

The initial logical explanation is that the missing 20% items have demand that is not satisfied, thus sales are lost.  If those 20% would be available they will be sold according to their natural demand.

The claim is shown in a simple chart:

The right hand-side represents the original claim, then some more explanations on current lost sales that would not be lost now.  The oval shape says that the two causes act together.

Two different reservations to the above logic are:

‘Some customers might buy the same item somewhere else.”   And: ‘Customers might buy another item instead of the missing item.’  Both reservations aim at the causal arrow connecting unavailability of items to losing sales, from that effect, together with the improvement, to the resulting effect of ‘Sales go up”.

The two reservations highlight a clarity issue. The improvement cause is stated “We improve…”, but who are ‘we’?   It could be the management of the chain of stores, the local management of a particular store or a supplier of a family of items.  Each of them gives a different meaning to the current state and then has its own reservation to the claimed effect of “Sales go up”.  The supplier of certain products means ‘his products’ are available only in 80% of the time and customers who buy replacement products cause the supplier to lose sales. If the availability of the supplier products would go up, then those specific products will be sold more.

This is a non-trivial ‘clarity’ issue.   We first have to deal with the clarity reservation by making a choice.  I have chosen the perspective of the store, and now I have to relate to the causality reservation doubting whether unavailability of an item always causes loss of sales to the store.

When customers don’t find a specific item they might buy a similar item.  In this case the store does not lose the sale.  In other cases the clients might simply give up.  In some rare cases the client might walk out, which could mean other sales are lost as well.  So, we conclude that some sales are lost because of unavailability, but the direct loss of sales is less than the calculated average sales of that item in the period of time it is short.

So, the above logical claim seems valid, but its real impact could be low.  We like to go deeper into the question when is the loss of sales due to unavailability significant?

Is the loss of sales equal for all items?

There are two parameters that make a significant impact on the loss of sales for the store when an item is missing.  The first parameter is the average level of daily sales and the second is the level of loyalty of the clients to the brand/item.

Fast runners, when they are short, create considerable damage not just to the direct loss of sales, but also to the reputation of the store – meaning customers might look for a different store in the future.  The logical statement is: if ‘a fast runner is missing’ then ‘many customers are pissed off’ causing ‘some regular customers look for another store to make their purchasing’ causing ‘total sales go significantly down”.  I’ve added ‘significantly’ to make a mark about the total impact.

But, as ‘management are aware of the potential damage to the store from missing fast-runners’ then we expect the following effect to apply: ‘management is focused on maintaining the perfect availability of fast-runners’.

So, we can deduce that if ‘the current management is reasonably capable’ then ‘the missing items do not include fast runners’.  Of course, 20% of the items being short might still mean non-negligible amount of sales of medium and slow movers being lost.   The open question is how much and even more:  how the current level of shortages impacts the reputation of the store and through this the future sales?

So, we need to look deeper into the impact of the second parameter – loyalty to a specific brand/item.  The effect ‘some items are special for some clients’ causes the effect that ‘some customers develop loyalty to that item’. This effect causes ‘the probability that some customers refuse to buy a replacement is high’.  Thus, if ‘items with strong loyalty are frequently missing’ then ‘some customers try other stores’.  The effect of ‘items with strong loyalty are frequently missing’ also causes ‘our reputation for what we carry on the shelves goes down’, with clear impact on the future sales.

The difficulty with ‘loyalty of customers to the brand/item’ is that it is difficult to validate its power.  The true test for the strength of loyalty is when the item is short and checking whether the sales of alternative items go up or not.

One additional reservation from the basic claim that improving the availability of items on the shelf would increase sales:  it assumes that ‘most customers entering the store know exactly what they want to buy’.  If this effect is not valid, then what is important for the sales is that the shelf is full with items that have good enough demand.  The effect that some items planned to be on the shelf are missing, but other items, with equal chance of being sold, fill the space well, would not cause a clear impact on the sales.  The kind of items that people come to browse and then choose (‘when I see it I’ll know’) have to managed in a very different way than maintaining availability of specific items.  For such items it makes sense to replenish them with new items, unless a specific item seems such a hit that maybe keeping it available is beneficial giving the high desirability of clients.

The effect of ‘The store has many regular customers’ has an impact on the meaning of ‘availability’ on the incidental customer. A shop in a big airport serves mostly incidental clients, so unavailability of items doesn’t impact future sales.  When there are no regular customers, then there is no difference between items that the store does not hold and items that are short.  This is relatively a small issue.

There are many more conditions that we consider true without further thought: ‘we live in a free economy’, ‘there are many competing choices for most items’ and ‘there is enough middle-class customers that can afford buying variety of products’.  If we try to include all ‘sufficiency’ conditions we’ll never end up with anything useful.  On the other hand it also opens the way to major mistakes due to hidden assumptions about what not to include in the analysis.  One needs the intuition when to stop the logical analysis, recognizing also the validity of ‘never to say I know’ (an insight by Dr. Goldratt).  Another aspect is the impact of uncertainty:  there are no 100% cause and effect relationships.  But, causal relationships that are 90%, or more, valid are still highly valuable.

Eventually we get the following structure as a summary of the above arguments.  Not all the previous effects have been mentioned, which means some of the logical arrows require more details, but eventually this is the claim.

We still cannot determine how much the sales would go up, because it depends on the characteristics of the medium and slow runners:  how many of them have strong loyalty.  If we add to the initial effects also ‘The chain makes marketing efforts to radiate the message that the chain maintains very high availability at every store’ then the chain can expect a faster and stronger increase in its reputation and in its sales.

Was it worth to go through logical analysis? 

While we still have only a partial picture, it is probably better than a picture based just on intuition without any analysis.