Before the Click: Direct Response Roots (1870s–1993)

New York, 1925. A first-year copywriter hands his boss a stack of headlines for a mail-order piano course. The boss reads for maybe a minute, checks one with a pencil, says write copy for that one. Meeting over.

The copywriter was John Caples, and the headline was “They Laughed When I Sat Down at the Piano… But When I Started to Play!” which is still getting ripped off a hundred years later. The headline isn’t the interesting part. The pencil is. A man picked one option out of several, ran it, and found out whether it worked, because in mail order a person always found out whether it worked — the order came back or it didn’t. That pencil is modern PPC in miniature.

Attribution ran on paper for a century

Chicago, 1872. Aaron Montgomery Ward mails a single-page price list of 163 items out of a rented shipping room, aimed at farmers getting gouged by rural general stores. By the mid-1890s that page had become a catalog pushing a thousand pages and roughly $7 million a year in goods. Sears did it bigger — founded 1886, renamed Sears, Roebuck and Co. in 1893 — and the mailing numbers stop being numbers and start being weather:

318,000 catalogs to the Midwest in 1897
3.6 million by 1908
more than 50 million a year by 1916

Richard Sears wrote much of the early copy himself. The founder was the copywriter, because the copy was the store.

Mail order is inherently measurable. The argument about whether advertising works never happened, because money arrived in the mail or it didn’t. And once advertisers started keying the coupons and order forms, coding each one so a return could be traced back to a specific publication, a specific headline, a specific offer, the industry had source-level attribution. Not a metaphor for it. The real thing, in 1890, in ink.

Hopkins was running split tests in 1923

Claude C. Hopkins got hired by Albert Lasker at Lord & Thomas in Chicago (1907 or 1908, the sources fight about it) for a reported $185,000 a year, which clears $5 million in today’s money. For a copywriter.

He earned it by treating advertising as arithmetic. Scientific Advertising, 1923, about 95 pages, opens with the line everybody quotes: “Advertising is salesmanship.” The other one is the one that lands harder, where he writes that mail-order advertising gets traced down to the fraction of a penny, and that the cost per reply and the cost per dollar of sale show up with exactness.

Cost per reply. Cost per dollar of sale. CPL and ROAS, described in 1923, described as solved, described as the boring part that the rest of the industry refuses to do because it would rather talk about art. His method was a loop: key the coupon, trace the return, run headlines and offers against each other, keep the winner, kill the loser, run it again. I’ve read enough modern testing frameworks to say this plainly. That’s A/B testing, fully formed, with the vocabulary missing.

It worked on real accounts. His Schlitz campaign walked readers through the brewing process step by step — the sterilized bottles and the filtered air and the deep wells — and reportedly moved the brand from around fifth to tied-for-first (that specific claim gets repeated everywhere and is hard to pin down, so hold it loosely). He built Pepsodent by dramatizing the film on people’s teeth, which is why anybody brushes every morning at all — Charles Duhigg gives him direct credit in The Power of Habit. Palmolive, Quaker Oats, Van Camp’s, Goodyear. Same loop every time.

Caples formalized it. An engineer by training, Naval Academy, first year at Ruthrauff & Ryan when he wrote the piano ad. He’d drafted alternatives, including one that opened by asking whether the reader could play piano, and it lost to social humiliation and vindication, because it always loses to social humiliation and vindication. He wrote Tested Advertising Methods in 1932. The thesis is in the title. Tested. Not great.

The Yellow Pages were paid search with worse latency

Reuben H. Donnelley created the first official Yellow Pages directory in 1886. (The Cheyenne printer who supposedly ran out of white paper in 1883 is folklore, since Donnelley would have been about nineteen and living in Chicago, so file it under tall tale and move on.) The model: categorized listings, free to the consumer, funded entirely by advertisers, bigger and more prominent ads costing more. Bill Gross said outright that GoTo.com’s keyword bidding was inspired by the Yellow Pages, where businesses pay more for larger ads.

Nobody ever opened the Yellow Pages for entertainment. It got opened because a water heater was leaking at nine at night, and the medium captured demand that already existed rather than manufacturing it, and routed that demand to whoever had bought the most prominence in the category. Strip out the paper and that describes Google Search exactly. Same psychology, same auction, same economics. The only real difference is that the directory updated once a year and the auction updates every time somebody types.

US print peaked somewhere around $13–15 billion in the early-to-mid 2000s ($13.6 billion in 2001 is the firmest anchor). By 2015 it was roughly $3.2 billion, and it sits near $1.1 billion now, with print distribution falling from 540 million directories in 2007 to about 75 million. More than ninety percent, gone. The demand didn’t go anywhere. It moved to a search bar.

A roofer in Fort Worth buying search ads this morning is running Claude Hopkins’s playbook on Reuben Donnelley’s medium. It only feels new because nobody told him it’s a hundred and fifty years old. — The History of PPC, Part 1

What this actually means

Split testing is Hopkins, 1923. Headline testing as a discipline is Caples, 1932. Cost per lead and cost per acquisition are Hopkins again, in a book that takes an hour to read. Paying for prominence is Donnelley, 1886. Capturing existing intent instead of creating it is the entire Yellow Pages business, running profitably for a century before anyone bought a keyword.

The channel changed. The feedback loop shrank from weeks to minutes and the auction got smart enough to price itself, and the delivery got faster than anyone in 1923 could have modeled, and none of that touched the underlying math or the underlying psychology.

Carryover: the vocabulary that survived

Terms established in this era and the instruments they became. Names changed. Functions did not.

Keyed coupon · 1890s: A code printed on an order form so a returned sale could be traced to the publication, headline, or offer that produced it. UTM parameters, gclid, tracking templates, dynamic call-tracking numbers.
Cost per reply · Hopkins, 1923: Total spend divided by responses generated. Cost per lead (CPL), cost per acquisition (CPA), target CPA bidding.
Cost per dollar of sale · Hopkins, 1923: Ad spend expressed as a ratio against revenue produced. ROAS, target ROAS bidding.
Split test · Hopkins, 1923: Two versions of an offer run against each other under identical conditions, winner retained. A/B test, ad variation experiment, campaign experiments.
Headline testing · Caples, 1925–32: Systematic comparison of opening lines against measured response, body copy held constant. Responsive search ads, asset performance ratings, pinned headlines.
Reason-why copy · Hopkins, Schlitz: Persuasion built on specific mechanical proof rather than assertion of quality. Landing page experience as a scored Quality Score component.
Category listing · Donnelley, 1886: Classification of a business under the term a buyer would look under. Keyword, search query, match type.
Paid prominence · Yellow Pages: Position and visual weight sold at tiered rates, larger placement priced higher. Bid, Ad Rank, absolute top impression share, asset extensions.
Intent capture · Yellow Pages: Advertising placed where demand already exists rather than where it must be created. Search vs. Demand Gen, bottom-funnel budget allocation.
The offer · mail order, universal: The specific proposition and terms under which the transaction occurs. Conversion action, primary CTA, promotion extension.