How DigiCom Diagnoses Stalled Meta Accounts in the Andromeda Era

Hey everyone,

Welcome back for another bite to chew on.

Last year's Meta playbook does not work in 2026. Andromeda was supposed to make life easier for marketers. Instead, it has made the work 10 times harder, and most brands have not caught up to the change.

DigiCom is an 11-year performance marketing agency on the Inc 5000 that audits Meta accounts for a living. They work across eight verticals, from CPG to health and wellness to kids, pet, and apparel. Monthly client spend ranges from $20K all the way up to $1.5M, which means they see the same patterns at every scale.

Across those audits, three failures keep showing up. None of them are about budget. They are about reach, attribution, and creative diversity. Get any one of them wrong and your scale flatlines. Get all three wrong, and you spend the next quarter wondering why the algorithm has turned on you. Here is the diagnostic.

On the Menu:

Most stalled Meta accounts are not broken at the budget level. They are broken at the reach, attribution, and creative-diversity level. And the team running the account is usually too close to see it.

That is where an outside set of eyes earns its keep. DigiCom has spent 11 years running this exact diagnostic across hundreds of DTC brands, from $20K monthly spenders up to $1.5M.

Get a free Meta audit from DigiCom →

The number one trap Jacob sees in audits is frequency. Specifically, frequency on the wrong audiences. Most accounts run frequencies above 10 on engaged and existing customer audiences over the course of a week. They are spending the bulk of their budget reminding people who already know the brand to buy.

Meta's algorithm is happy to do this. The people most likely to convert next are the ones already close to converting. So if you do not tell Meta otherwise, it will spend down your retargeting audiences and starve the top of your funnel. The numbers look efficient. The growth stops.

"Oftentimes when we break this down, we look at an account. People have frequencies on their engaged and existing audiences that are above 10 over a course of a week," Jacob said. "They're overserving that bottom of the funnel."

The fix starts with a metric called net new reach. Take your current month's total Meta reach, subtract the prior month's, and divide by the current month. That percentage is how many of your impressions reached people who had not seen you the month before.

DigiCom's target is 70%. The audit account they walked through was running at 57% on $800K of worldwide spend. That means 43% of impressions went to people who had already seen an ad in March. On a worldwide targeting setup with that kind of budget, 57% is a top-of-funnel collapse.

"That should be closer to 30%, even 20% for some of our accounts," Jacob said about the retargeting share. "We don't want to overserve this retargeting warm audiences."

The mechanical fix is aggressive exclusion. Past purchasers. Site visitors. Facebook and Instagram engagers. Your CRM list. Your email list. All of it. Excluded from any campaign that is supposed to be prospecting.

This goes against the instinct most operators have, which is to let Meta find the most efficient conversion. But Meta's definition of efficient is "the person about to buy anyway." If you want incremental customers, you have to take those people off the table and force the algorithm to look for new ones.

The second move is defining your audience segments in Meta's ad settings: engaged, existing customer, and new. Until you do this, your reporting collapses every audience into "unknown," and you cannot see which group your frequency is actually concentrated in.

What you can do: Calculate your net new reach this month versus last. If it is under 60%, your prospecting campaigns are not actually prospecting, and your scale will plateau before your budget does.

Picture a customer who clicks Ad 1 on Monday. Then Ad 2 on Tuesday. Then Ad 3 on Thursday, and finally buys. Meta's default attribution gives all the credit to Ad 3. Ads 1 and 2 show zero purchases. On paper, they look like losers.

So you pause them. You optimize toward Ad 3. And now you only have a bottom-of-funnel converter running. You wonder why no new people are entering your funnel. The answer is you turned off the introduction ads because they did not look efficient under a measurement system that cannot see them.

"Ad 1 and Ad 2 will show zero purchases even though they contributed, potentially, and I would argue more, to driving the purchase because they introduced the new person to the brand," Jacob said. "But Meta's not going to give them any credit."

DigiCom's standard account structure now runs three campaigns instead of two. An ABO testing campaign with prospecting-only audiences and broad creative tests. A CBO top-of-funnel scaling campaign optimized to first-touch attribution on a cost cap. And a CBO bottom-of-funnel scaling campaign optimized to last touch on a target ROAS, with Advantage+ retargeting.

Each scaling campaign answers a different question. The top-of-funnel asks: which ads bring new people into the brand at the right cost? The bottom-of-funnel asks: which ads close the deal once they are warm? Different ads will rise in each.

The key move is you stop letting last-touch CPA decide which ads survive. Top-of-funnel winners are evaluated on a 28-day or lifetime first-touch window. Bottom-of-funnel winners stay on the standard seven-day click.

If you have a third-party attribution tool like Triple Whale or Northbeam, this is where you earn the investment back. Sort your ads by the delta between first-touch CPA and last-touch CPA. Ads with low first-touch CPAs and higher last-touch CPAs are your introduction ads, the ones that bring people in. Push them into the top-of-funnel campaign.

Ads with low last-touch CPAs and weaker first-touch numbers are your closers. They are great at converting warm traffic. Push them into the bottom-of-funnel campaign. Ads that win on both go in both.

If you do not have third-party attribution, Meta's incremental attribution setting is the next best option. It signals you care about the most meaningful touch in the journey, not just the last click. It is a black box, but it is directionally better than last touch alone.

What you can do: Pull the last 60 days of your Meta ads into Triple Whale or Northbeam and sort by first-touch CPA. The ads with low first-touch and high last-touch numbers are the introduction ads you have probably been pausing.

Andromeda does not stop at the headline anymore. It listens to your videos. It analyzes lighting, background, facial expressions, voiceover, music, and setting. Every element of an ad becomes a signal the algorithm uses to match it to a viewer.

This is contextual targeting. If a user is currently watching dog videos on Instagram, Meta is more likely to serve them an ad that visually reads as a dog ad. Interest audiences are not just dead. They are obsolete. The creative itself is the targeting.

The implication is brutal for most accounts. If your ad library is full of variations on the same visual recipe, the algorithm has no signal to work with. It cannot find new audiences because every ad looks like the one before it.

The trap John walked through was creative iteration that looks diverse but is not. Same product shot, same background, three different hook texts. To a human reviewer, that is three ads. To Andromeda, it is one ad with three labels.

"With the Andromeda update, the way that Andromeda is reading this is it is labeling all three of these ads, even though they are different, slightly different. It is labeling it with the same entity ID," John said. The result is you think you are testing three concepts. Meta thinks you are serving the same ad three times.

Real diversity changes the visual framing, the first two to three seconds, the audio hook, the text hook, and the persona. Keep the concept consistent, but change enough of the DNA that the algorithm treats them as separate ads.

DigiCom's working method now is three to five personas per brand, each with a different creative recipe. In the example John walked through, an Italian charm bracelet brand built three. The sentimental gifter (a 44-year-old buying for family). The trend adopter (a 21-year-old building a wrist stack for social). The identity expressor (a 35-year-old looking for symbolism).

Each persona gets its own creative ecosystem. The gifter gets testimonials from gift receivers and unboxing reactions. The trend adopter gets short-form video with aggressive hook testing in the first two seconds. The identity expressor gets carousels and four-panel designs that show off catalog variety. That is how you generate genuinely different visual DNA at volume, instead of 20 variations of the same idea.

The connective tissue is naming conventions. DigiCom tags every ad with format, type, color, hook style, header, and landing page destination. That is how you pivot the data and answer questions like whether UGC is outperforming polished video by 20%. Or whether blonde influencers convert better than brunettes in a kitchen setting versus a gym. Without the labels, the data is noise. With them, you close the door on what is not working and pull on the thread of what is.

What you can do: Audit your last 20 ads. If you can swap the hook text and the rest looks identical, Andromeda is treating them as one entity ID, not three tests.