Statistics that make you go duh? Connections that make you go hmm
- Summary:
- There are some weird correlations out there but some bear further examination.
How does the old saying go: "Lies, damned lies, and statistics" - well how about this example above? The correlation may not be terribly close but on its face seems close enough for someone, somewhere to get worried about watching Nick Cage movies while lounging near the backyard pool.
Or how about the inverse relationship described below?
I'm sure some scientists somewhere will have found this strong correlation to be true but you'll note that inverse relationship didn't really break out until the late 90's.
Mostly these are of course spurious correlations but as the amounts of data to hand start to come under scrutiny then I have to wonder what other pearls will be revealed. This is not a laughing matter although I have to admit to smirking at some of the examples shown at Spurious Correlations.
Over at the Kellogg School of Management,they're looking closely at The Customers You Do Not Want. This is a fascinating and well researched topic that looks at the possible connection between failed products and customers who loved the product - so called 'harbingers of failure.' If that sounds whacky then read what the researchers discovered:
...researchers found that just 40 percent of new products are still in stores three years later, a number in line with previous estimates. But critically, a product’s chances of succeeding depend not only on how much is sold but also on who is buying. The surprising finding is that when sales increase to a segment of consumers whom the authors label “harbingers of failure,” then the new product is more likely to fail.
This seems counterintuitive and may yet turn out to be one of those classic CPG false positives. Even so, the researchers say:
The study suggests some rather simple approaches market researchers can take to weed out niche products before they ever appear in stores. Most obviously, firms should ask customers not just whether they would purchase the product in question, but what other products they regularly purchase. A customer who regularly purchases Swiffer mops, for instance, probably has fairly mainstream tastes and should be paid attention to.
But the customer still pining after that unusually flavored beer that was only in stores for three weeks? Market researchers should definitely pay attention to them.
Quite how these relationships work is not well understood and may only apply in certain markets and product types. Even so, as the compute power available to researchers increases, we can expect to see a proliferation of models that claim to unearth all sorts of unexpected correlations with attendant cause and effect theories tagging along. (sic)
But as always in these things, the ability to stand back and apply a healthy dose of initial skepticism or at least be prepared to dig further should always be considered.
Images via Spurious Correlations.