fm.

Meta ads account structure: why I cut 20+ ad sets down to three to five

The Meta account I ran at Shopify had 20+ ad sets split by audience. After iOS 14.5 that structure stopped learning. This is the order I rebuilt it in, what happened to CPA and how the same thinking applies to a small local account.

Key takeaways

  • Run three to five broad ad sets in place of dozens of small audience splits. An ad set that gets too few conversions can't learn.
  • Fix the signal before the structure. Send conversions from your server through the Conversions API first.
  • With broad audiences, creative does the targeting. Consolidation only works if creative volume goes up with it.
  • At Shopify this rebuild held CPA within 15% of pre-ATT levels after iOS 14.5, while much of the industry swung 30% to 50%.
  • Keep the source of truth outside the ad platform. Meta's numbers are an input.

01What a Meta ads account structure should look like

A Meta ads account structure that holds up is three to five broad ad sets fed by server-side conversion events, with many creatives inside each one. Splitting an account into dozens of small audience ad sets starves each of them of the conversions it needs to learn.

I didn't start there. The Meta account I ran at Shopify had 20+ ad sets split by audience before iOS 14.5. It had lots of audiences and lots of control, and it looked sophisticated. It was also fragile, and Apple's privacy change showed how fragile.

02Why an over-segmented account breaks

Meta's delivery learns ad set by ad set, from the conversions each one records. Twenty ad sets split the same pool of conversions twenty ways. While the pixel saw nearly everything, that was survivable. When iOS 14.5 cut the signal, three things happened at once:

  • Small ad sets stopped getting enough conversions to learn from.
  • Attribution windows changed.
  • Platform CPA became less reliable as a number to steer by.

Every extra split makes the first problem worse. The fix was fewer ad sets, a better signal and a different idea of where targeting comes from.

03How I rebuilt it, in order

  1. Server-side events first.With the martech and data teams, I led the move to the Conversions API and cleaner first-party events in about 30 days. That recovered roughly 25% of the conversion signal we'd lost. There's no point restructuring around events the platform can't see.
  2. Consolidate the structure.20+ segmented ad sets became three to five broad ones, so each had enough volume to learn.
  3. Let creative do the targeting.A broad audience only works when the ads themselves sort people. Creative volume and structured testing went up at the same time. The method is in my note on an ad creative testing framework.
  4. Move the source of truth out of the platform.Meta and Google numbers became inputs. Budget decisions ran on our own merchant dashboard and on holdout tests.
BeforeAfter
Ad sets20+, split by audienceThree to five, broad
Conversion signalBrowser pixelConversions API and cleaner first-party events
What does the targetingAudience definitionsCreative
Creative volumeSpread thin across ad setsHigher, with structured tests
Where decisions get madePlatform CPAOur own merchant dashboard and holdout tests

I'm not going to tell you how to divide your three to five ad sets, because that depends on your funnel. The check is the same for everyone: look at how many conversions each ad set records in a week. If several are starved, merge them.

04What it did at Shopify

CPA held within 15% of pre-ATT levels while much of the industry swung 30% to 50%. The Conversions API work recovered roughly 25% of the lost signal, and the consolidation meant the signal that remained was concentrated in ad sets that could use it.

The structure also depended on what the ad sets were optimizing to. When Shopify rolled out its $1 a month trial, I moved bidding and reporting from signups to paid conversions on Google and Meta. Signup CPA rose about 20% and cost per paying merchant fell about 30%. A tidy structure pointed at a cheap event still buys the wrong people. The setup for that is in my note on value-based bidding.

05The same idea in a small account

Consolidation applies at the account level too. At Dental Team FL, a group of ten dental practices, five agencies each ran their own channel and didn't share data. I moved everything into one program with one Meta ad account, one set of targets and one report.

What worked on Meta there was simple:

  • One master ad account for all ten practices.
  • Plain static image ads, which brought the lowest cost per lead.
  • Consultation campaigns pointed at implants and Invisalign.

Health ads have rules, and some of ours couldn't mention pricing. When Meta rejected ads, I reworked the creative with compliance and relaunched.

06Mistakes I've made or seen

  • Over-segmenting for the feeling of control. This one was mine. When the signal dropped, too many ad sets had too little volume to learn.
  • Consolidating without adding creative. Broad ad sets with the same few ads give the system nothing to sort people with.
  • Restructuring before fixing events. Do the server-side work first, then decide how many ad sets the signal can support.
  • Steering by platform CPA alone. After a signal loss it's the least reliable number you have. Check it against your own customer data.

Rebuilding a paid social account around the customers who pay is work I do as a performance marketing consultant.

FAQQuestions people ask

What is the best Meta ads account structure?
A small number of broad ad sets fed by server-side conversion events, with plenty of creative inside each one. At Shopify, Farhan Munshi cut a Meta account from 20+ audience-split ad sets to three to five on the Conversions API after iOS 14.5. CPA held within 15% of pre-ATT levels while much of the industry swung 30% to 50%.
How many ad sets should a Meta ads account have?
Few enough that each one gets the conversions it needs to learn. Farhan Munshi's Meta account at Shopify went from 20+ segmented ad sets to three to five broad ones. The exact number matters less than the check behind it: look at how many conversions each ad set records in a week, and merge the ones that are starved.
How do you restructure Meta ads after iOS 14.5?
In this order: server-side events, then structure, then creative, then measurement. At Shopify, Farhan Munshi led the move to the Conversions API in about 30 days, which recovered roughly 25% of the lost conversion signal. He then consolidated 20+ ad sets into three to five, raised creative volume and moved budget decisions to an internal dashboard and holdout tests.
Should I use broad targeting on Meta?
Broad targeting works when creative does the sorting, so it needs more creative than a segmented account does. When Farhan Munshi consolidated Shopify's Meta account into three to five broad ad sets, creative volume and structured testing went up at the same time. Going broad with the same handful of ads gives the system less to work with.
Should a multi-location business use one Meta ad account or several?
One, with one set of targets and one report. At Dental Team FL, a group of ten dental practices, Farhan Munshi replaced five separate agencies with one centrally run program and ran Meta from a single master ad account. Plain static image ads brought the lowest cost per lead, and consultation campaigns pointed at implants and Invisalign.