Stop picking the audience. The ad does the whole job now.

01What changed, in plain words

Meta changed how it picks who sees what in the middle of 2026. The ad that wins now is plain and made for 1 group. Straight problem, straight pain, their whole day. Nice lighting became optional. Being about somebody did not.

Narrow the audience now and you are telling a very good system to ignore people it would have found.

The machine got good at the job we did by hand. It cannot know what your buyer says about their problem at 11pm. So button pushing is worth little now. The scarce skill is knowing which groups of buyers exist and saying their problem back to them. A listening job, not a settings job.

02The whole structure fits in 5 lines

Here is all of it, on 1 screen. That is part of the point.

The common mistake is in the first line. If you want booked calls, do not switch to link clicks because the cost per click looks nicer. You get cheap clicks from people who never book.

Before strangers, spend on the people who came to your page and left. You already paid to get them there, so it is the cheapest money in the account. Put the pixel of every platform you might advertise on into your pages, not only today's. This does not fight broad targeting. Broad finds new people. Following old visitors finishes the job, and doing that badly is what makes broad look expensive.

03The funnel is not a setting in your account

Everything above is the half you can see. The half that decides it is which level of knowing each ad talks to. There is no box for that in the account.

These 5 levels came first and the top, middle and bottom words came later. Same idea, new clothes. So 15 ads spread across 15 groups and across these levels.

If all 15 talk to people ready to buy today, you built 15 bottom of funnel ads. You burn that small pool and everything falls apart in 2 weeks. People call it ad fatigue. Usually the ad is fine and the pool is empty.

04Make 15 openings, not 1 perfect ad

1 ad talks to 1 version of your buyer, and a service business never has 1 version. A physio clinic talks to a runner with a bad knee, a woman in her 40s with desk back pain, a patient out of surgery, and a parent who cannot lift their own kid. 4 conversations.

1 ad that tries to hold 4 buyers holds none of them.

So you make 4. Then 8. Then 15. Not 15 offers, 15 doors into the same offer. We keep it cheap by recording the body once, the part that explains the offer, and keeping it the same for everybody. Only the first 5 to 7 seconds change.

One opening is a call this week with a founder who wanted 1 dashboard. The next is a director in Dubai with a renewals problem. After those 7 seconds, the same recording. A new group costs 1 opening, not 1 ad. 15 tests a month instead of 3.

The opening has to be real and from this week. We keep the week's calls written down in our command center, so the lines come from what somebody actually said. Make them up and the ad goes flat. Name the group, say their situation in their words, then the offer.

05Give it 10 days, then move 20 percent

Give a test about 10 days before you judge it. Not 3. On a small budget, 3 days of numbers is noise in a costume, and killing an ad on day 3 is how owners end up sure nothing works.

10 days also matches how you really sell. For most service businesses the sale does not happen in the ad. It happens on a call 2 weeks later. Judge on same day sales and you have misread your business.

The rhythm is boring too. Every week, 2 or 3 new ads from real research, not a brainstorm. They go live overnight, so a clean day starts in the morning, next to the ads already running, in the same campaign. A new test never starts a campaign from zero.

Decide your number first. Pick the cost per real result you need to make money, write it on 1 line, and judge every ad against it. Without it you compare ads to each other and feel good about the least bad one.

When the account beats that number, put the budget up by about 20 percent and leave it alone. Not double. Big jumps throw the campaign back into learning and you pay twice. And 1 expensive day is not a signal. Sitting still is a real decision, usually the right one.

06The number you trust, and what ads cannot fix

Read cost per real result, angle by angle. For a local service business that is cost per booked appointment. Not cost per click, and not cost per lead if a lead is an email address. And not the platform's own report card.

Those are our numbers, not a benchmark. Compare your ads to each other, never to somebody else's screenshot. And the dashboard is not the scoreboard. Its window is short, so a sale that closes in 2 weeks gets less credit than it earned. The phone that saw the ad and the laptop that booked look like strangers.

The greedy part costs the most. Somebody sees 5 of your ads in a week and all the credit goes to the last click. So the opener gets killed, and kill enough openers and the closer has nobody left. Keep 1 number of your own, booked calls or closed deals, and check the platform against it.

How hard should it be to become a lead? Your process does not make good leads, it sorts for them. A video that must be watched, a real application, a phone number you have to give: those raise quality. A slow page only sorts for patient people. Too many bad leads, make it harder. Too few, make it easier. Nothing in the way and still bad leads, then it is your targeting or your offer.

This structure will not fix a weak offer. Ads only help more people say no faster. It will not fix slow follow up, and I have watched owners blame the ads for the 26 hours before anybody called back. It will not fix being full, because more leads become a longer queue. It will not fix a budget too small to learn. And it does not fix being invisible where nobody buys ads, because plenty of buyers now ask a chatbot who to hire, which I wrote about in will a chatbot say your name.

07Questions, answered