The Squawk Point

Organisational Mechanics

  • Home
  • Blog
    • People
    • Data
    • Process
    • Wild Cards
    • Index
  • Podcast
  • Book

What is Sample Bias?

5 January, 2015 by James Lawther 12 Comments

There are some jobs you really wouldn’t want…

Perhaps the most dangerous in history was to be a crew member of a World War II bomber (I don’t suppose it matters which side).

You shouldn’t believe everything you read on the internet, but according to some of the more reliable sources, during World War II:

  • Over 12,000 Bomber Command aircraft were shot down
  • 55,500 aircrew died.
  • The life expectancy of a Lancaster bomber was 3 weeks
  • Tail-gunners were lucky if they survived four missions.

A terrible waste of lives…

But it was insisted that bombing was critical to the success of the war, so all this destruction lead to a very pressing question:

How do you cut the number of casualties?

Arguments raged:

  • One line of thought was the bombers should be more heavily armoured so reducing the damage wrought by the anti-aircraft batteries and enemy fighter planes
  • Another was that they should be lightly armoured so they could fly quickly and avoid the lethal bullets in the first place.

What was the solution?

The Hungarian-born mathematician Abraham Wald solved the problem by plotting the bullet holes found on battle-scarred planes on a diagram. It looked something like this:

What is sample bias? An example is returning bombers in World War 2. The sample is based in favour of those that survived.

The wings and fuselage presented the biggest target and were most likely to be hit, so those were the areas that would most benefit from being armour-plated.

A fairly straightforward deduction to make.

But Abraham Wald disagreed. His belief was exactly the opposite, the armour plating should be placed in the areas that hadn’t been hit.

He argued that the sample was biased.

What is sample bias?

Any sampling method that systematically favors some outcomes over others is biased. All the analysis was carried out on planes that returned, not those that were shot down.

Being hit in the areas that were undamaged was far more likely to prove fatal. The unmarked areas were the ones in need of armour plating.

A tragic story but why tell it?

It highlights a number of points

  • Data is more useful than speculation
  • The Pareto principle holds true — some things are always more important than others.
  • Visual analysis (plotting where things have gone wrong) is a useful ploy.
  • You should always worry about sample bias.

But perhaps the most important point…

Always apply a little common sense — hits to the cabin and engine are less dangerous than the fuselage — really?

If you enjoyed this post click here to have the next delivered straight to your inbox


Read another opinion

Filed Under: Blog, Operations Analysis, Tools & Techniques Tagged With: assumptions, aviation, data is not information, pareto principle, sample bias, six sigma, video, visual management, World War 2

About the Author

James Lawther
James Lawther

James Lawther is a middle-aged, middle manager.

To reach this highly elevated position he has worked in numerous industries, from supermarket retailing to tax collecting.  He has had several operational roles, including running the night shift in a frozen pea packing factory and carrying out operational research for a credit card company.

As you can see from his C.V. he has either a wealth of experience or is incapable of holding down a job.  If the latter is true this post isn’t worth a minute of your attention.

Unfortunately, the only way to find out is to read it and decide for yourself.

www.squawkpoint.com/

Comments

  1. john.allen says

    6 January, 2015 at 11:11 pm

    True, I still fly as a crew-chief on a B-25 plane. The old quote the bullet comes in one way goes out the other-no protection. – the pareto test is good. You have to remember the bulk of most armor was around the pilot.

    Reply
    • James Lawther says

      11 January, 2015 at 8:25 am

      Thanks for the comment John

      Reply
  2. Adrian Swinscoe says

    7 January, 2015 at 4:17 pm

    James,
    I think the same logic could be applied to complaints – those that are received/given and those that are not.

    Adrian

    Reply
  3. John Hunter says

    17 January, 2015 at 5:44 am

    I first heard example from George Box years ago. It has stayed with me ever since. Using data requires thinking. It is often hard to get people to use data more in the first place. But even when you do, the way it is used still leaves a great deal to be desired. Examples like this help people realize that critical thinking is necessary to understand what the data is showing you.

    Reply
    • James Lawther says

      22 January, 2015 at 9:43 pm

      Thanks for your comment John, we do appear to be wired to jump to conclusions rather than think things through

      Reply
  4. Vance Butler says

    19 January, 2015 at 5:40 pm

    To Comment on John Hunter’s point – people need to realize there is a difference between data and information. They are not the same. Not mutually-exclusive either, you need both, one informs the other. Sometimes it can be hard to get users of data to seperate the two.

    Reply
    • James Lawther says

      22 January, 2015 at 9:44 pm

      Yes Vance, and I often find that for many the solution is simply more data, thanks for your comment

      Reply
  5. maz iqbal says

    19 January, 2015 at 10:41 pm

    Hello James,

    Excellent illustration of the biases in our thinking. And how real thought takes being with the challenge-question longer than most of us. Or put differently, sitting in the question rather than jumping in with the first answer. Some say, that our natural wiring, is not designed for this kind of analytical thinking.

    It occurs to me that here it is worth listening to Karl Popper. Specifically, operating from the perspective-stand that all knowledge is provisional: not true so much as not yet shown to be false. And that the scientific way to deal with any hypothesis is to seek to refute it, not to prove it.

    All the best
    maz

    Reply
    • James Lawther says

      22 January, 2015 at 9:46 pm

      I like the idea of all knowledge being provisional Maz, that is very astute.

      Reply
  6. David Cary says

    22 October, 2017 at 8:49 pm

    I am finishing a book. As lead in to how we often misinterpret data, I reference the Wald story.

    I would like to use the image you used. It sounded like you might have created it. If so, may I use it in my book?

    If not, could you tell me where you got it?

    Thanks,

    Reply
    • James Lawther says

      23 October, 2017 at 8:45 am

      I mocked it up David, there are other versions on the net but I didn’t want to infringe anybody’s copyright. Help yourself.

      Good luck with the book, let me know when you publish.

      JL

      Reply

Trackbacks

  1. The World’s Most Dangerous Job? | James Lawther | Michel Baudin's Blog says:
    16 January, 2015 at 2:40 pm

    […] Source: https://www.squawkpoint.com […]

    Reply

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Explore

accountability assumptions beliefs best practice blame bureaucracy capability clarity command and control communication complexity continuous improvement cost saving culture customer focus data is not information decisions employee performance measures empowerment error proofing fessing up gemba human nature incentives information technology innovation key performance indicators learning management style measurement motivation performance management poor service process control purpose reinforcing behaviour service design silo management systems thinking targets teamwork test and learn trust video waste

Receive Posts by e-Mail

Get the next post delivered straight to your inbox

Creative Commons

This information from The Squawk Point is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.
Creative Commons Licence
Customer Experience Update

Try This:

  • Fish Bone Diagrams – Helpful or Not?

  • Regression to The Mean

  • Should You Punish Mistakes?

  • The Alpha Male Paradox

Connect

  • E-mail
  • LinkedIn
  • RSS
  • YouTube
  • Cookies
  • Contact Me

Copyright © 2025 · Enterprise Pro on Genesis Framework · WordPress · Log in