The platform picks who sees your ad

What I do

The full round: from whether the ad account hears about your enquiries to switching off an ad that is not working.

• 01

  Make the ad account aware of your enquiries

  Every enquiry from the site reaches the ad account. It sounds dull and it decides everything: an ad account that never sees an enquiry starts bringing you people who only tap the picture.

• 02

  Look at what competitors are running

  Their ads are visible to everyone. The ones that have run for months are paying for themselves, or they would have been pulled. It is a free hint almost nobody uses.

• 03

  Collect reasons to buy

  Every ad answers one question: why would this person buy from you. You need many reasons and they have to differ. Twenty ads about the same thing count as one.

• 04

  Make many different ads

  One reason becomes a dozen versions in an evening: a different opening line, a different picture, a different length. That used to be a week of filming.

• 05

  Launch, and switch off by rules

  Few campaigns, a wide audience, many different ads inside one of them. The rules for switching an ad off are written before launch, not at the point where you have spent enough on it to want to keep it alive.

• 06

  Connect what happens after the click

  The enquiry lands in one place and does not get lost between a messenger, an inbox and somebody's memory. This is my main line of work, and advertising without it pours water into a leaking bucket.

What changed, and why the old instructions fail

Three shifts that rewrote the job.

• The platform now chooses, not the media buyer

  You used to describe an audience and the platform looked for people inside it. Now it looks at the ad itself and decides who should see it. The interest checkboxes are still there, but narrowing an audience usually gets in the way rather than helping.

• What counts is not how many ads but how different they are

  The platform treats similar ads as one and makes them compete with each other. Twenty versions of one idea work like one. Ten different ideas work like ten.

How the work goes

The first step happens once. The rest repeat in rounds, and each round gives the next one its material.

1. 1

   Agree what counts as a result

   An enquiry, a call, a booking, a purchase. Then I make sure the ad account learns about each one straight away. The site reports it directly, bypassing the browser: browsers block half of those reports, which leaves the advertising blind to half of your enquiries.

2. 2

   Study the market and the competitors

   I look at which of their ads have been running for a long time, because an ad nobody pulled is paying for itself. I read the reviews and comments their buyers leave. That is where the words people use for their own problem live, and they almost never match the words a company uses for its product.

3. 3

   Write down the reasons to buy

   I turn the findings into a list: why someone buys from you. Not ads, not headlines — reasons. You want around a dozen visibly different ones, or there is nothing to choose between.

4. 4

   Make the ads

   Each reason becomes a dozen versions: a different opening line, a different picture, a different length. This is where AI works, and it is the one place it speeds the job up several times over.

5. 5

   Launch

   A few campaigns with a shared budget and a wide audience, holding all the different ads at once. The budget has to be enough for the advertising to learn: it needs roughly fifty enquiries a week. On less than that it learns forever and never quite works.

6. 6

   Switch off what fails and count the result

   The rules for switching off are written in advance. Then I reconcile what the ad account reports with what you see in your own enquiries, because the account tends to claim credit it did not earn. At serious budgets I check separately, by switching the advertising off and watching what sags.

What the AI does here, and what it does not

A split that saves money and temper.

• Does: turns one idea into many executions

  Opening lines, copy variants, storyboards, backgrounds, voiceover, cut-downs for different formats. What used to be a week of production is an evening.

• Does not: know your customer

  A model will write persuasive copy about a benefit you do not have. The reasons people buy come from reviews, from the people who take the calls, and from your own head. They are not invented.

Measurement: three numbers instead of one

The ad account tends to claim credit it did not earn; end-to-end analytics tends to lose credit it did. The truth sits between them.

• The account's number

  Useful for steering bids and switching ads off. Useless for deciding whether the channel as a whole is working.

• The business number

  Everything spent on advertising against everything earned in the period. Crude, but it does not lie and does not depend on an attribution model.

• The experiment's number

  The only way to learn the lift: compare against a group that saw no ads, or honestly switch the channel off and watch what sags. Worth doing once budgets are serious.

When I turn this work down

The limits, stated plainly, so your money is not wasted.

• While a lead has nowhere to land

  If leads fall between a messenger, an inbox and somebody's memory, advertising increases the losses rather than the revenue. Intake first, traffic second.

• While the advertising cannot see your enquiries

  With nothing reported back, the platform optimises for what it can see: clicks and views. That is a different job with a different outcome.

• Below the learning threshold

  If the weekly budget cannot produce enough target events to gather evidence, the campaign lives permanently in learning. Better to narrow to one offer and one city than to spread thin.