World War II was in full swing with the Allied War Room at high alert. The Allies were suffering huge losses at the hands of the Axis Powers. But more than anything else, the General heading the War Room was troubled by the loss of bomber planes to enemy fire. The action plan he had in mind was to add armor to the body of the bombers. But there was a challenge. Due to limited resources, he could add armor to only selected parts of the plane, not the entire body. But which parts of the plane, should he prioritize?
He called one of his team members, an Analyst expert in Military Strategy.
"But I don't know anything about aircraft," replied the Analyst when the General explained the problem to him.
"You don't need to," answered the General. "Just go through the records we have and analyse the pattern. Come up with a solution to this."
The Analyst was still not convinced but he knew very well not to pester the General anymore who was in a hurry to leave for another meeting.
"What's the name of that mathematics professor," asked the General from the other end of the hall. "The one who is working on cracking Enigma. He will help you out with the records."
Before the Analyst could answer, the General had left.
"I need the data of all the damaged bomber planes we have along with the details of their damaged areas," announced the Analyst, reaching the professor's desk.
"Why," asked the professor without looking up. He was working on some kind of a miniature machine. The Analyst knew the professor was within his rights to question him. He handled a repository of sensitive information, after all. Still, the Analyst was a bit irritated by the question.
"So that I can analyse the damage pattern and accordingly plan to armor those specific parts of the planes only."
"Ok... " The Professor took a long pause. "So you need records of all the planes or just the damaged ones."
"Obviously, just the damaged ones," replied the Analyst.
The Professor slowly put his machine aside and rested his hands on the desk in a thinking position.
"Don't you think your approach is wrong?" It was a rhetorical question because he kept speaking. "The damaged planes are the ones that could take damage and still fly well enough to return to base. If you analyse the pattern of bullet holes in them, it will give you what to exclude, not the other way around. I think, once you have analysed the damage pattern, you should put armor on the parts that are not damaged because the planes hit in those areas are most likely to be lost in battle."
Someone snapped his fingers.
"Vishal, I need the list of Active Dealers of Trade Sales along with the details of their Monthly Counter Potential (MT) and Month-wise Sales (MT) during this Financial Year."
"Why," replied Vishal, recovering from his daydream.
"In the Monthly Review Meeting, I have to present the reasons for Sales Decline in certain areas."
"Ok... what do you mean by Active? Do you mean the ones currently active in the system with a Credit Limit running? Or the ones who have non-zero sales volume in all the months of this FY? Or should I just consider the ones who have had non-zero sales in the last three months? Even then, what about the ones who have Sales Volume at the beginning of the FY but are currently inactive in the system? Just considering the Sales Trend of Active Dealers won't help you. You should also consider, even target the list of Dealers who have stopped working for us."
The Analyst gave a reply but Vishal could not hear it over the sound of the bombs and bullets.
During World War II, the statistician Abraham Wald took survivorship bias into his calculations when considering how to minimize bomber losses to enemy fire. He examined the damage done to aircraft that had returned from missions and recommended adding armor to the areas that showed the least damage. The bullet holes in the returning aircraft represented areas where a bomber could take damage and still fly well enough to return safely to base. Therefore, Wald proposed that the Navy reinforce areas where the returning aircraft were unscathed, inferring that planes hit in those areas were the ones most likely to be lost.
Just like Wald, while analysing data, one should avoid survivorship bias by considering what data points may be missing from a dataset and using accurate data sources that do not omit key observations.
***
Liked what you read? Leave a comment below.
Comments
Post a Comment