ABO vs CBO for lead generation: which should you use?

Use whichever one is already working for you. If neither is, use ABO: give every new ad its own ad set and budget, and judge it over four days. Then spend your energy launching more ads, because the setting matters far less than the ads.

I get asked ABO or CBO more than almost anything else. People expect a setting. What they need is a way of testing and scaling that doesn't sabotage itself.

Why the setting matters less than you think

Most people treat the budget setting as the lever. They switch from one to the other when results dip, and nothing really changes. From my observation, that's because the budget setting was never the problem.

The ad is the lever. What it says, who it's clearly for, and whether it gives the right person a reason to keep watching. Get that right and the budget setting barely matters. Get it wrong and no setting saves it. I make the same point about broad versus interests in this video on asking the wrong questions.

When I tested what ChatGPT does with this question, it went straight to Meta's help pages on budget settings. The pages it cited explained the mechanics well. None of the ones I read showed results measured past the form.

How I set up testing

Each new ad gets its own ad set with its own budget. My floor is about $50 AUD a day per ad set. That's budget at the ad set level, so if you need a label, it's ABO. But the label isn't the point.

The point is that every ad gets a fair shot. Your total budget decides how many ads you can test at the same time. It does not decide whether ads work.

If your budget only covers one or two ad sets, test one or two ads at a time. Don't spread the same money across ten ads and call it a test.

Judge over four days, not daily

Daily numbers bounce around. People see a bad day, panic, and edit the ad, the audience or the budget. Every small edit resets learning, and then they wonder why nothing settles.

The tinkering is the problem. I judge each ad over four-day windows and leave it alone in between. I explain why four days in this video.

And I judge on what a customer costs, not what a lead costs. A lead that costs more but books and buys is cheaper than a lead that never answers the phone.

How I scale without restructuring

This is where most accounts go wrong. Someone finds a winner, builds a new "scaling campaign", moves the ad across, and it falls over. Leave winners where they are.

• Vertical. Change the budget on a winner where it sits, up or down, by no more than about 20% in a day.
• Horizontal. Keep launching and proving new ads. No single ad carries a campaign forever.
• Both, when the budget allows.
• Duplicate, never edit. If you want a variation of a winner, duplicate it. Editing a winner throws away what it has learned.

Here's what that looks like with example numbers. Say an ad has held a cost per customer of $280 over two four-day windows at $50 a day. I raise that ad set to $60 and leave it for four days. If cost per customer holds, it goes to $72, then about $86, and I pull it back a step if cost per customer climbs. Alongside it, I keep launching two or three new ads at $50 a day each, so there's always a next winner coming through.

What matters more than the budget setting

• The objective. For lead generation, use a lead form or conversion objective. No awareness funnels until lead gen is already profitable.
• The signal. On conversion campaigns, set up the Conversions API so your ad account receives what happens after the form.
• The targeting. Go broad. The ad does the targeting, as long as it's specific about who it's for.
• The form. Qualifying questions with conditional logic, so the wrong people can't submit.

FAQ

Should I use ABO or CBO for lead ads?

Use the one that is working for you. If neither is, use ABO: give every new ad its own ad set with its own budget so it gets a fair test, then put your time into launching more ads.

How much budget does each test need?

My floor is about $50 AUD a day per ad set, with one ad per ad set. Your total budget decides how many ads you can test at once, not whether ads work.

How long should I run a new ad before judging it?

Four days. Judge over four-day windows, never daily. Editing the ad, audience or budget every day resets learning and makes the results meaningless.

Should I move winning ads into a scaling campaign?

No. Leave winners where they sit. Raise or lower the budget by no more than about 20% in a day, duplicate rather than edit, and keep launching new ads alongside them.