Targeting got automated. Message did not.

For about a decade, media buying skill meant targeting skill. Stacked interest audiences, lookalike percentages, exclusion pyramids. Agencies sold that complexity as the craft, and for a while it was.

Then the platforms took it in-house. Meta’s own best-practice guidance now tells advertisers to hand delivery broad audiences and let the system find buyers, with creative variety as the main input the advertiser still controls. The machine got the targeting job. What it cannot do is decide what your ad says to the person it finds. That job did not get automated, and it is now where almost all of the remaining edge lives.

The research said message was the edge all along

In 2017, Nielsen Catalina Solutions published a meta-analysis of nearly 500 campaigns measuring which parts of advertising actually drove sales. Creative - what the ad says and shows - contributed about 47 per cent of the sales impact, and around 56 per cent for digital campaigns specifically. Targeting contributed about 9 per cent.

Read those two numbers again, because most ad accounts are organised as if they were reversed. Endless audience experiments, one set of creative run into the ground. The largest measured lever, by a factor of five, is the message. The study was consumer packaged goods, so treat the exact percentages as directional for services rather than gospel. The ordering, though, matches everything we see in call-booking funnels: when an account turns around, it is almost never because the audience changed. It is because the message finally matched a pocket.

What “better creative” actually means

This is where the advice industry goes vague. “Make better creative” usually decays into make more creative, or make louder creative. Volume and polish are not the variable. Specificity is.

A century ago Claude Hopkins wrote in Scientific Advertising that platitudes and generalities roll off the human understanding like water from a duck, and that specifics sell. Nothing about the feed has changed that. The ad that stops a clinic owner is not the one with the best production values. It is the one that opens on the thing that happened at her front desk this morning. We wrote about where that specificity comes from in the audience pockets piece: it is mined from how buyers in one situation actually talk, then written for that pocket and no one else.

The division of labour that works

The account structure that follows from this is simple to say:

• Let the platform do the finding. Broad delivery, minimal constraints. Fighting the machine with audience micro-slicing is paying 2015 prices for a job that is now nearly free.
• Do the saying yourself. One pocket, one message, so the ad itself becomes the targeting: it stops the right people and slides past the wrong ones, and the delivery system learns from that signal who to find next.
• Judge the pair together. A pocket-matched ad under broad delivery tells you something clean: this message, in front of the people it stopped, produced calls that closed or did not. That is learnable. A generic ad under narrow targeting tells you nothing.

The uncomfortable implication

If message is the remaining edge, then the scarce skill in paid acquisition is no longer inside the ads manager. It is knowing the market well enough to say something specific to each pocket of it. That is research and writing, not button configuration - which is why an account can be technically perfect and still dead. The settings were never the ceiling. The sentence at the top of the ad was.