Advertisers Were Randomly Testing Ads Before Doctors Were
Mixed press runs meant neighbors could get different ads by chance, so sellers could count which one sold.
My First Million · Rory Sutherland: completely legal marketing hacks that always workMail-order sellers were letting chance decide which ad each home received, then counting which one sold more, by the 1920s, possibly earlier. That was decades before medicine widely used the same kind of chance-based test, now called a randomized controlled trial.
These were advertisers who sold directly through coupons clipped from papers and codes quoted over the phone, so each sale could be traced to a specific ad. They tried out headlines in tiny, cheap classified ads before paying for a costly full-page ad.
They used the way newspapers were printed. Papers ran on two or sometimes four presses at once, and the copies were mixed together on the way out. One version went in copies from one press and a different version in copies from another, so neighbors could get different ads by chance. It is unclear who arranged the setup.
That avoided the bias if wealthy areas got one ad and poorer areas got another. In some cases, advertisers might find one paper sold four times better than another, or one ad version sold twice as well as the next, from very small changes.
Questions people ask
When did medicine start using randomized controlled trials?
The first famous modern randomized trial was the British Medical Research Council's 1948 test of streptomycin for tuberculosis. Statisticians like R.A. Fisher had developed randomization in agriculture in the 1920s, but it took decades to become the required standard for approving drugs.
Do advertisers still test this way?
Yes, and more aggressively. Email marketers routinely split lists to test subject lines, and websites and apps run A/B tests that show different headlines, prices, or buttons to random visitors. The newspaper press trick was an early physical version of what software now does automatically.