Note to parents: Washing your kid’s mouth with soap doesn’t prevent cussing. I know this because when my son was small, he tested it. Upon seeing a repeat-offender friend’s mouth cleansed for the umpteenth time, he came home, applied soap to his own mouth and then tried to say “damn.” And succeeded.

Little did he know that he’d used the scientific method. And if potty-mouthed kids can use it, so can marketers who want to increase their effectiveness.

First, let’s explain the scientific method in “Real People-ese:” You start with a hunch (soaping a mouth prevents cussing); you do a predictive test (if that’s so, then I shouldn’t be able to swear after soaping my mouth); and you draw a conclusion from the results (I could still say “damn,” so the hunch was most likely wrong).

Consider, for instance, how I used this method to help a client choose between two covers for a romantic music CD. One cover was black with a red heart in the middle. The other featured a golden sunset. In intercept interviews, nearly everyone scoffed at the sunset cliché and expressed a strong preference for the sexier black cover.

I wasn’t convinced. As a direct marketer, I know that arguably trite elements like time-worn phrases, gaudy design elements and peel-off stickers can pack a good deal of selling power. For a more predictive test, I stacked the two versions side-by-side on a tray, told people that the CDs were identical and offered to let them take one free. I did not reveal that I was testing, and I didn’t ask for opinions. Each time someone chose a CD, I replaced it to keep the stacks equal. I also switched the position of the stacks in case people were biased toward one side or the other. The results were eye-opening. Though nearly everyone had expressed a preference for the black cover, when it came time to choose one, nearly everyone opted for the sunset. Cliché and all.

The more I resort to the scientific method, the more I’m surprised at what I learn. I must admit that, in the above-referenced example, both the client and I preferred — and were rooting for — the losing design. Other tests have equally shown that I am a fallible clairvoyant. I was certain that a free offer for an entertainment center remote control would outperform a free offer for a five-inch flashlight. But in a series of direct mail tests, the flashlight won three times over.

In a direct mail test for a university fundraiser, I “knew” that a mission-centered appeal would outperform a member-benefits appeal. But when we tested and observed, there was no difference in response. Further testing revealed that response rose only when the school’s football team won a significant tournament. (Not an easy strategy to roll out, as fixing games isn’t one of my agency’s core capabilities.)

The facts aren’t always what you want them to be, which is why the scientific method is important. Careful testing, observing and tallying are the surest ways to protect against seeing what you hope to see at the expense of facts.

Sadly, not all marketers want facts. I know many who readily disqualify all inconvenient data. Still others exhibit remarkable creativity when it comes to making data turn out the way they want. One enterprising fellow split a test mailing 90-10 instead of 50-50 to ensure that the version he preferred resulted in a higher gross response. In his report, he hid the fact that the other version outperformed on a percentage basis.

Someone should have washed his mouth with soap.