A WWII RAF fighter plane on the ground... two officers inspecting bullet holes clustered in the tail section... the fuselage punctured and scarred, RAF roundel visible, dramatic overcast sky... the planes that made it back, carrying the wrong lesson

The Importance of Perspective and Intelligent Analysis

During World War II, numerous fighter planes were getting hit by anti-aircraft guns. Air Force officers wanted to add some protective armour/shield to the planes. The question was "where"? The planes could only support few more kilos of weight. A group of mathematicians and engineers were called for a short consulting project.

Fighter planes returning from missions were analysed for bullet holes per square foot. They found 1.93 bullet holes/sq. foot near the tail of the planes whereas only 1.11 bullet holes/sq. foot close to the engine.

Tail Section
1.93
bullet holes / sq. foot
Air Force conclusion: armour here
Engine Area
1.11
bullet holes / sq. foot
Wald's conclusion: armour here

The Air Force officers thought that since the tail portion had the greatest density of bullets, that would be the logical location for putting an anti-bullet shield.

A mathematician named Abraham Wald said exactly the opposite; more protection is needed where the bullet holes aren't - that is -around the engines.

The insight
Abraham Wald, Mathematician
"We are counting the planes that returned from a mission. Planes with lots of bullet holes in the engine did not return at all."

Debrief

If you go to the recovery room at the hospital, you'll see a lot more people with bullet holes in their legs than people with bullet holes in their chests. That's not because people don't get shot in the chest; it's because the people who get shot in the chest don't recover.

What you don't see in the data is often more important than what you do.

How analysis in detail actually helps?

We have been working on a technology to automatically identify road conditions (still in development phase) and while testing there was one particular stretch of road that completely baffled us for some time.

Our tech is built in a manner that it should should be able to detect potholes and speed breakers on the road automatically using sensors. However, on one stretch of road (which is complete ridden with potholes - to an extent that you can say that the road does not exist), our technology would never gives us an alert about the condition.

We had tested over 1000 kms and we were getting data with more than 90% accuracy, but this 100 meters stretch kept eluding us.

A Mahindra Bolero barely crawling through a completely destroyed Indian road... more craters and mud than surface... people wading alongside at the same pace... the road that ceased to exist, moving at 5 km/h
The road was so bad that anyone travelling on the patch could hardly move at more than 5 kilometres an hour. Our technology categorised this movement as walking and hence ignored all sensor results.
The answer hidden in plain sight
One a deeper analysis, we realised that the road conditions were so bad, that anyone travelling on the patch could hardly move at more than 5 kilometres an hour. As the speed was so low, our technology categorised this movement as walking and hence ignored all sensor results! Pretty simple answer to a query that could not be answered by analysis more than 1 million data points.

Why lack of data is also an important part of data analysis?

Being in the service industry and consulting clients for more than a decade now, I have been used to hearing the phrase 'We need data analysis to take decisions'. However, a lot of times, we get into the 'data' itself so much that 'analysis' takes a back seat. First-hand understanding is often what no amount of data can replicate.

Therefore, it is imperative that while analysing situations, you do not just look at the information itself, but the way information was collected, the sources used and the conditions in which information was collected. One a wholesome understanding, can give a result that is is valid.

📚
Story source
From the book... "How Not To Be Wrong" by Jordan Ellenberg
← All writing Home