Goodness-Of-Fit

Research analysts are expensive. That’s why, in lieu pf paying six figs for some smarmy analyst with a fancy-pants degree, we’ve decided to just hire our six-year-old niece, Paprika. She’s a pretty smart kid; if anyone can take the pulse of the automobile customization industry in which we work, it’s her.

Paprika’s first mission is to find out which of our existing products are most appealing to the general public. After six weeks and several thousand dollars in research and travel costs, Paprika informs us that, by and large, our most appealing product is…hot pink. Not window tint, not low-profile rims, not even upgraded stereo equipment, but hot pink. Confused by her findings, we ask her to walk us through her research study, and she shrugs and tells us it’s simple: she asked her ten best friends (also first-graders) what they would most want on a car of their own. And overwhelmingly, their answer was the same: hot pink.

Friends, this is what we call a bad research study. Why in the world did asking ten first graders one question cost several thousand dollars and take six weeks?

But more importantly, (and more germane to our discussion here today) this study absolutely and unequivocally fails the goodness-of-fit test. A goodness-of-fit test is when we look at a research study done on a sample group and determine to what extent the results found can be extrapolated to represent the general population. What Paprika should have done if she’d been thinking straight (which clearly she wasn’t, and we’ll need to talk to her mother about that), is to set up a more specific survey that would be administered to a random representative sample of the general car-owning population. That would’ve been a lot more useful, since we could reasonably assume that those results would be more indicative of how our entire potential customer base feels. In other words, that study would’ve resulted in a better goodness of fit. And (probably) almost no one would have replied with, “hot pink.”