In 1966, The Fugitive aired an episode every marketer interested in AI should probably watch.
A scientist unveils the 2130, a “digital computer” that studies Richard Kimble’s movements, spots his habits, and predicts where he will go next.
And it works. Lieutenant Gerard starts tightening the net.
Then Kimble discovers the machine. So he changes up his behavior.
Game over.
The computer cannot catch the man who refuses to keep acting like the man in its files.
Claude Hopkins understood that advertising should be judged by what buyers actually do. Not what the copywriter likes. Not what the client prefers. Not what six people in a conference room vote for.
Run the ad, count the response, let the loser die.
Every test gives you another scrap of intelligence about the buyer. Stack enough of those scraps together and your control becomes a prediction: this message, aimed at these people, should produce the most money.
Until it doesn’t. Because markets refuse to sit still for your spreadsheet.
Competitors steal your angles. Customers become harder to impress. Fresh hooks rot from repetition. Yesterday’s killer appeal gets beaten into mush.
Testing discovers what the market is rewarding now.
A control is useful only as long as reality keeps cooperating.
Keep testing, or eventually your customer becomes Richard Kimble and walks straight out of your model.
RW
.
.
.