[The Best of Reskilling in the Digital Age] Logical Thinking Part 2: Cause, Effect, and the Convenience Store Urban Legend

*If you haven't seen Logical Thinking Part 1, click here

Causal relationship is the relationship between cause and effect and it means explaining the reason properly. Humans are creatures that will not be convinced unless they understand the reason behind something. Children repeat "Why? Why?" from a young age, and they instinctively know that if they can get someone to accept a reason, such as "Everyone has a Switch, so please buy me one," it will lead to action.

Furthermore, there is a rule for elementary school essay writing called the Iritama rule. It is an acronym for "Opinion," "Reason," "Experience," and "Summary", and it is said that if you write in this order, you can write a highly persuasive essay. The "reason" appears second, and
the importance of explaining the "reason" properly is also asserted in essays where logicality is tested. It should also be noted that the "opinion" and "experience" in "Iritama" have an equal relationship between abstract and concrete. It is quite a good rule.

As a term with a meaning close to causal relationship, there is correlation. The proper use of these two terms is important from two perspectives: improving logical thinking skills and fundamentally understanding investment.

The relationship between causal relationship and correlation is shown in a Venn diagram at the bottom of the figure below.

As you can see from this, causal relationship is completely contained within correlation. The important points suggested by this are the following two.

1. There are areas where correlation holds but causal relationship does not 2.
Correlation is a necessary condition for a causal relationship to hold

1 is the cause of confusion in business settings due to misinterpretation. Let me give a concrete example. This is one of the convenience store urban legends.

A convenience store manager: "Sales of meat buns are good today, so the number of customers will be tough."
A clerk at that store: "Indeed, looking at past records, on days when meat buns are selling well, the number of customers is invariably low."
Manager: "Meat buns are the culprit that drops the number of customers. I'm sick of it. Maybe I should stop stocking meat buns."
Clerk: "It would make operations easier, so I agree!"

What is strange about this conversation? There is a negative correlation between meat bun sales and the number of customers. That much is OK.

So, is selling meat buns the culprit (=cause) for the decrease in the number of customers? This is where it is wrong. If that were really the case, it would be impossible for all convenience stores to continue stocking meat buns.

Then, what can explain the negative correlation between meat bun sales and the number of customers? In this case, it is thought that there is a separate cause, and that cause is influencing meat bun sales and the decrease in customers respectively. For example, temperature. When the temperature is low, meat buns sell. When the temperature is low, it is troublesome to go outside, so the number of customers decreases. A correlation has arisen between the results that were influenced by the same cause, which is temperature.

Besides this, a common one is the sales-request problem. There is a correlation between the number of sales staff and sales, and the claim is that our department has few sales staff, so sales are not going up, so please increase the number of sales staff. It is the logic that the number of sales staff is the cause and sales are the result. If I were the management team, I would ask, "If increasing sales staff definitely increases sales, I would be happy to increase them, but is that true?" "First, try increasing sales. Then I will increase the sales staff." This would be the natural way of thinking for the management team.

This is a case where the direction of cause and effect is reversed. If you write a time-series graph, in this case, the fluctuation in sales occurs ahead of the fluctuation in the number of sales staff.

Referring to the examples above, I want you to be aware not to easily overinterpret correlation as a causal relationship just because it exists.

The point of discussion here is, then how can you tell if there is a causal relationship? To conclude, it is to satisfy the following three points. *Let A be the result and B be the cause

1. There is a correlation between A and B.
2. There is a temporal sequence between A and B. A changes after B changes.
3. Things considered to be causes other than B are unrelated to A. Unrelated means there is no correlation or no temporal sequence.

The problem is 3. Although it is written casually, if you read it carefully, it asserts something that is practically impossible to prove

There are various causes. For example, in the case of the convenience store urban legend mentioned above, as things that affect meat bun sales, besides temperature, there are infinite candidates for causes, such as an event happening nearby by chance, meat buns being sold out at the convenience store next door, or someone planning a meat bun party buying them in bulk. You have to prove one by one that all of these are unrelated. That is probably impossible.

If we keep saying things like this, the conversation won't progress, so in everyday conversation, I think it's best to proceed with the assumption that something is the cause if there is a certain degree of probability.

However, business doesn't work that way. The foundation of business is taking action against a cause in order to produce results, and it is usually not acceptable to say after taking action that there was actually no causal relationship.

So, is there a way to strictly prove a causal relationship? The answer is YES. That is A/B testing. A/B testing is a method of verifying by preparing two randomly sampled groups, changing only the condition for which you want to prove causality between the groups, and actually conducting a test. It is often used in web marketing, for example, in cases where only the color of a banner is changed to verify the causal relationship between the color and the result of the click-through rate.

However, in real-world business, you cannot always perform A/B testing for everything. When A/B testing is not possible, after doing your best to pursue the cause, you will ultimately have to take the risk yourself and assert the causal relationship. This courage after doing your best is extremely important.

Click here to proceed to the next section ↓