When analyzing any individual data and drawing conclusions from it, we often run the risk of making the typical "Survivorship bias".

What is it? Let's get to the bottom of it

‍

Survivorship bias is false conclusions based only on known facts without taking into account those that are unknown.

A prime example of Survivorship bias and its solution is the work of Hungarian mathematician. Abraham Wald, who worked for the American army during World War II.

Wald was tasked by his command to analyze bullet and shrapnel holes in American planes and propose a way to armor them so that pilots and planes would not be killed.

Solid armor could not be used - the airplane would be too heavy. It was necessary either to armor those places where there was damage, where bullets hit, or those places where there was no damage. Wald's opponents suggested armoring the damaged areas of the returned aircraft (marked with red dots in the picture).

‍

Wald objected. He admonished that we should analyze the damage to those, planes that were shot down and could not return. Since this would mean that their damage affected more important points.

Wald's suggestion prevailed. Aircraft were booked where the returning machines had no damage. As a result, the number of surviving airplanes increased significantly. By some accounts, Wald saved the lives of about 30% of American pilots in this way.

‍

Returned airplanes 

❌ Do not armor 

Airplanes that did not return

✅ Armor

‍

"Systematic Survivorship bias"

It's easy to make in digital marketing. By analyzing individual data, you risk missing something important and failing to see the big picture.

That's why you should always work COMPLETELY with data

‍