Examples of surprising data

April 29, 2012 12:23 PM

What are some citable examples in which companies or people might reasonably expect one outcome but find something different or surprising upon actually examining the data?

I'm looking for examples that show something surprising along the path from "reasonably expected" to "actually observed," like these:

• A canned soup manufacturer removed cooking instructions during the label redesign for their "heat and eat" soup. Subsequent consumer testing found that a few people put the whole can in the microwave (I can't find a reference for this one now).
• The 2010 US census shows the percentage of parenting among same-sex couples is highest in the South.
• Canadian Tire found that those who used their credit cards to buy generic automotive oil were far more likely to miss payments than those who bought carbon-monoxide detectors.
• The four countries with the largest Muslim populations are neither predominantly Arab nor in the Middle East.

posted by tomwheeler to Society & Culture (10 answers total)

The spiral galaxy NGC4622 rotates in the opposite direction from expected.

These results were met with skepticism in part because they contradicted conventional wisdom with one quote being “so you’re the backward astronomers who found the backward galaxy.”

posted by xil at 12:33 PM on April 29, 2012

I don't know if it's what you're looking for, but the New Coke fiasco seems like a good example based on what I think you're asking about.
posted by Golfhaus at 12:53 PM on April 29, 2012

In web design, a/b testing often yields unintuitive or very surprising results. Like how an ass ugly ad made in MS Paint results in a better conversion rate than a professionally looking ad. Because of this, a/b testing tends to make web developers/designers uncomfortable because you realize that in many situations you cannot blindly trust your gut feeling.
posted by Foci for Analysis at 1:09 PM on April 29, 2012

Depends on who you ask, but many people might find it surprising that teen pregnancy and divorce rates tend to be highest in conservative red states, and lowest in liberal blue states. (This isn't really because the states are liberal/conservative, but might be the opposite correlation than is expected).
posted by nakedmolerats at 1:51 PM on April 29, 2012

Malcolm Gladwell's book Blink has a lot about this, including an interesting exposition of the New Coke story that Golfhaus mentioned.
posted by spbmp at 1:57 PM on April 29, 2012

Fareed Zakaria said something this morning about how Americans believe we have hundreds of thousands of illegal immigrants coming in from Mexico but so many are going back (because of our bad economy?) that right now we are at -zero net gain-.
posted by cda at 3:04 PM on April 29, 2012

I don't have a link, sorry, but several studies have shown that people who take daily multivitamins are MORE likely than others to die of various causes such as heart disease and certain cancers.

Also, I think that diet soda has been correlated with weight gain/obesity as much as, or even more than, regular soda? Something about the artificial sweeteners disrupting the body's ability to signal you to stop eating when you're full? (I'm very cloudy on the details of that one but it's something vaguely along those lines.)
posted by désoeuvrée at 3:22 PM on April 29, 2012

The quality of school you go to makes much less of a difference to your attainment level in education than what a lot of people think - only about 10% - the actual main factors are your socio-economic background and your parents' attitude to their own education. (Recently heard a radio documentary that touched on this)

There's also a lot of results in psychology... like visualising a successful outcome is actually demotivating, you are better off visualising the process to achieve a successful outcome rather than the result

Also like the 'Barry Manilow t-shirt experiment' as a nice meta example
posted by fearfulsymmetry at 3:40 PM on April 29, 2012

Benford's Law
posted by penguinicity at 4:38 PM on April 29, 2012

This is the very definition of apocryphal, but about 10 years ago a friend of mine was doing some data mining for a chain of service stations/mini supermarkets. They wanted to know what products people bought together to make more attractive promotions. So if lots of people bought both bread and milk, run a special on milk and make it up with the extra bread sales, or whatever.

He said that one of the strongest pairings for products were porn mags and condoms. I have yet to see an advertised promotion for porn and condoms.
posted by AndrewStephens at 6:20 PM on April 29, 2012