Chapter 23

Advertising

An ad has one job: to tie your name to a moment in the buyer's memory.

What changed

Advertising is the build most firms judge by what happens next. Clicks, cost per lead, return on ad spend, sales the dashboard says the ad caused. Each one is counted within days of the ad running.

That habit rewards the wrong ads. Most people who see an ad aren't buying, and won't be for months. Dawes's rule of thumb, that about 95% of business buyers aren't buying at any one time, is in Chapter 2. An ad judged on this week's sales gets bent towards the few who were about to buy anyway.

Those are the buyers you were most likely to get without it. Switching off eBay's paid search in 68 regions made no measurable difference to sales (Blake, Nosko and Tadelis). So this chapter treats an ad as a memory, written now and collected later.

What it is for

For nearly everyone who sees it, an ad's work is to leave your name filed under their moment. That's Known.

Its second, smaller job is to put you in front of someone at the moment they look: a paid search ad, a top marketplace listing. That's Found, bought rather than earned.

I think of these as writing the memory and collecting on it. Brand advertising writes. Search ads, retargeting and offers collect. Les Binet and Peter Field studied 996 campaigns from the IPA's effectiveness databank. Their rule of thumb puts about 60% of the budget into brand building and 40% into activation (IPA). Treat the split as a starting point. What matters is that writing gets a large share, because collecting can't work on a memory nobody wrote.

A good ad also carries your reason, which helps at Preferred. But advertising rarely builds belief. A claim isn't more true because you paid to repeat it. Answered, Bought and Kept belong to other builds.

The anatomy of a good one

A good ad has two halves. The creative half decides what gets written into memory. The media half decides how many memories, and how often.

The moment, in the first seconds (Known). Start with the situation, not the product: the hungry afternoon, the tax letter, the wage sheet. Memory is filed under these moments.

The brand, early and distinctive (Known). One study pooled 50,000 viewers' responses to 25 television ads. The later the brand appeared, the less often viewers recognised it (Gerber and colleagues). Show a distinctive asset early, then again through the ad.

A feeling (Known · Preferred). Over three years or more, 43% of emotional campaigns in the IPA databank reported very large profit effects. Only 23% of rational ones did (IPA). Pick a feeling that belongs to the moment.

One thing to remember (Preferred). One, not three. Often it's your reason, in words someone else could repeat. "Every Monday by nine" works because it's easy to keep in mind, not because no rival could do it.

Proof, then where to find you (Believed · Found). Point at evidence others have given you, then close on a name plain enough to type or say to an assistant.

Broad reach (Known). Growth comes from people who rarely or never buy you, so reach them too, not just last week's visitors.

Continuity and share of voice (Known). Memory fades. Stay on air at a steady level all year, not in one burst.

The mistake most teams make

They trust the attributed sales. Brett Gordon and colleagues compared common measurement methods against 15 randomised experiments at Facebook (Gordon and colleagues). The common methods "often fail to accurately measure the true effect of advertising", mostly overstating it.

The ads closest to the sale look best on the dashboard, and are the most likely to be paying for buyers who were coming anyway. But don't swing to the other extreme. Andrey Simonov, Chris Nosko and Justin Rao found that where rivals bid on a brand's name and the brand stayed off, the rivals took 18% to 42% of the clicks (Simonov, Nosko and Rao). Check who else bids before you switch branded search off.

None of this means collecting ads fail. It means you can't tell from the report.

How you know it's working

Use what I call the measurement ladder. Clicks and reach sit on the diagnostic rung: they show the ads ran. Steer by the leading rung, starting with searches for your name: Binet found share of search sometimes ran a year ahead of market share (IPA). Decide budgets on the causal rung: a holdout, where you switch ads off in some regions or periods and compare.

The bookkeeper's advertising

The bookkeeper for cafés is an invented example. Her budget is small. She made one ad: Monday, 7am, a café owner opening up and wondering if last week's wages ate the takings. Her name and her one colour appeared in the first two seconds and again at the end. The line was "Your wage cost against takings, every Monday by nine".

She ran it in two of her four towns for eight weeks, and not in the other two. She counted enquiries and searches for her name in each.

Nothing moved. With four towns, a small effect could hide in the noise. But she couldn't afford to keep paying on faith, so she stopped the ads and moved the money to reviews. She plans to retest in a year. A test that says no is worth as much as one that says yes.

The full playbook. The Advertising Playbook, with the ad audit scorecard, the ad brief, the holdout test plan and the measurement ladder, is free at rocmode.com/playbooks.

Whose evidence this is. The IPA is a trade body for agencies. Two of the Facebook measurement authors worked at Facebook, which sells ads. Peer-reviewed sources: Gerber and colleagues; Simonov, Nosko and Rao; Gordon and colleagues; Blake, Nosko and Tadelis.