Meta Advantage+ Is Underperforming. Here's Why — and What to Actually Fix.

Meta has spent the last two years aggressively pushing Advantage+ as the future of its ad platform, and a lot of advertisers did the reasonable thing: they followed the recommendation, turned it on, and waited for the automation to work. For a meaningful share of them, results got worse — or at best, no better than the manual campaigns they replaced. The natural conclusion is that Advantage+ is the problem, and the natural next move is to turn it off.

That's almost always the wrong read. After managing north of $100M in Meta spend, I've watched this pattern repeat across dozens of accounts, and the issue is rarely the campaign type. Advantage+ is an optimization engine, and an optimization engine performs in direct proportion to what you feed it. When it underperforms, the campaign type isn't broken. The inputs are weak, and the system is faithfully amplifying that weakness at scale.

What Advantage+ Actually Is

Strip away the marketing and Advantage+ is this: automation that hands targeting, placement, and creative-delivery decisions — the ones advertisers used to make manually — over to Meta's machine learning systems. Underneath it sit two systems working together. Andromeda decides which ads are eligible to reach a given user, reading the actual content of your creative to make that call. GEM, Meta's Generative Ads Recommendation Model, learns from your conversion patterns and shapes what the system prioritizes next. Meta detailed Andromeda publicly in late 2024 and published GEM's technical foundations in November 2025.

The critical thing to understand is what this machinery is and isn't. It is not a performance multiplier that makes weak advertising work. It's an amplifier. Strong creative signals, clean conversion data, and an aligned funnel get scaled up. Weak signals, thin data, and a broken funnel get optimized poorly — at scale, faster, with more of your budget behind the mistake. Advantage+ doesn't fix your inputs. It commits harder to whatever they already are.

Why Advantage+ Underperforms for Some Accounts

Four specific reasons account for most of it. Not "it depends" — these four.

Weak creative signals. Advantage+ reads your creative content to decide who to show it to. If your ad doesn't communicate a specific problem, a specific audience, and a specific outcome clearly, the system has nothing sharp to match on, so it casts a wide net and delivers impressions to low-intent users. Vague creative produces vague targeting. And creative is the input that moves the needle most — Kantar's analysis of a 250,000-ad database found top-quality creative drives several times more profit than weak creative. The automation isn't misfiring — it's matching on the fuzzy signal you gave it.

Misaligned funnel. Andromeda can match your ad to exactly the right user, but if the landing page breaks the promise the ad made, the conversion still doesn't happen. Advantage+ then scales the mismatch rather than the performance, driving more qualified clicks into the same broken handoff. Automation makes an alignment problem bigger, not smaller.

Thin conversion data. The system needs conversion volume to learn from. Accounts generating fewer than roughly 30–50 purchases a week give GEM too little signal to build an accurate buyer model, so it effectively optimizes blind. There's no automation shortcut around insufficient data — the model can only learn from events it actually receives.

Audience overcrowding from your own campaigns. Running Advantage+ alongside manual campaigns targeting the same audiences creates internal auction competition. Your campaigns bid against each other, inflating your own costs. This one is common and entirely self-inflicted — two of your campaigns fighting over the same people, and you paying the premium for the fight.

The Andromeda and GEM Dimension

Here's the part that explains why weak inputs don't just fail to help — they actively compound.

Advantage+ is Andromeda and GEM in practice. GEM learns from your historical conversion patterns and uses them to build a model of who your buyer is, then keeps finding more people like that. Now follow the logic: if your historical conversions are low-volume, or low-quality — driven by broad audiences with weak post-click conversion — then GEM is learning from bad examples. It builds a poor buyer model, and it uses that poor model to go find more of the wrong people, whose weak conversion behavior then feeds back into the model. The system isn't malfunctioning. It's learning exactly what you taught it, and what you taught it was wrong.

That's why turning Advantage+ off and back on, or tweaking campaign settings, changes nothing durable. The setting isn't the variable. The quality of the signal feeding the model is the variable. Fix the signal and the same automation that was working against you starts working for you.

I'll mark the honest boundary here, because it matters: Meta has published that these systems exist and broadly how they're structured, but not the exact weighting, scoring, or optimization mechanics. Everything above is grounded in Meta's disclosures and consistent practitioner observation — not a claim to read the source code. Be skeptical of anyone who tells you they know Andromeda's precise formula. They don't.

What to Check Before You Turn It Off

This is the practical core. Before you conclude Advantage+ doesn't work for you, run this sequence — because in most cases the fix is here, not in the campaign type.

Check creative signal clarity. Look at each ad and ask: does it communicate one specific thing to one specific type of person? If you can't state, in a sentence, exactly which problem and which buyer the ad speaks to, neither can the system. Broad creative is the single most common weak signal.

Check the landing page. Does the page immediately honor the promise the ad made — same claim, same audience, same words, above the fold? If the ad and page are selling different things, Advantage+ is scaling a handoff that leaks.

Check event match quality in Events Manager. Is CAPI feeding clean, deduplicated purchase data back to Meta? Poor match quality means GEM is learning from a corrupted picture of who bought, and every downstream decision inherits that corruption.

Check weekly purchase volume. Is the account generating enough conversion events to exit the learning phase and give the model something to work with? If you're well under 30–50 purchases a week, thin data is likely your real problem, and no setting fixes that.

Check for audience overlap with manual campaigns. Are you running Advantage+ and manual campaigns against the same people at the same time? If so, you're bidding against yourself. Separate the audiences or consolidate.

Four of these five have nothing to do with the campaign type and everything to do with the inputs. That ratio is the whole point.

A Worked Example: The Self-Cannibalization Trap

The audience-overlap cause is the one people find hardest to believe, because it doesn't feel like anyone did anything wrong. So here's the mechanism with numbers, since it's the most invisible and the most self-inflicted of the four.

An ecommerce account is running three manual campaigns it's had live for a year — a prospecting campaign, a retargeting campaign, and a broad "catch-all" campaign — spending about $15,000 a month combined and performing acceptably. On Meta's recommendation, the team adds an Advantage+ Shopping campaign with another $5,000 and leaves everything else running, because why turn off campaigns that work. Within two weeks, overall account ROAS is down, and the Advantage+ campaign specifically looks like a disappointment. The obvious read: Advantage+ doesn't work for us.

Here's what actually happened. Meta's ad auction is a single auction. When your Advantage+ campaign and your existing broad and retargeting campaigns are all eligible to show ads to the same high-intent past visitor, they don't cooperate — they enter the same auction as separate bidders and drive up the price of reaching that person. You win the impression either way, so you were always going to pay for it, but now you're paying a premium set by your own second campaign's bid. The auction can't tell that both bidders are you. It just sees demand for that user go up, and prices accordingly.

The damage concentrates on your best, most convertible audience — the warm, high-intent pool that every one of your campaigns wants to reach — which is exactly the traffic that was carrying your ROAS. So the blended number drops, and because the newest campaign is the one you're scrutinizing, Advantage+ takes the blame for a cost inflation that all the campaigns share. Meanwhile the Advantage+ campaign's own conversion data is being polluted: it's converting some of the same warm buyers your retargeting campaign would have converted anyway, so GEM is learning from overlapping, double-counted signal rather than clean incremental performance.

The fix isn't turning Advantage+ off. It's removing the overlap — excluding recent site visitors and existing-customer audiences from the Advantage+ campaign so it hunts genuinely new people, or consolidating the redundant manual campaigns into it rather than running both in parallel. Same budget, same campaign type, but now your campaigns stop bidding against each other and GEM gets clean signal about who converts incrementally. Nothing about the automation changed. The inputs stopped fighting.

When Advantage+ Is Actually the Wrong Tool

Being honest cuts both ways — sometimes the answer really is "not this account, not yet." Name the conditions clearly.

If weekly purchase volume is genuinely low — below roughly 20–30 conversions — the system won't exit the learning phase and will never optimize properly, no matter how clean your creative and funnel are. There simply isn't enough signal. And if your product has a complex or narrow buyer profile that doesn't emerge from creative signals alone — a niche B2B tool, a considered high-ticket purchase with a long research cycle — manual targeting may still outperform, because you know something about your buyer that the creative doesn't yet encode. In those cases, Advantage+ isn't failing you; it's the wrong instrument for the job, and forcing it is its own mistake.

When Advantage+ Works Well

And the other side, equally honestly. Accounts with strong creative libraries, clean conversion data, and healthy conversion volume often see real gains from Advantage+ — Meta reports that advertisers who turned on Advantage+ creative's AI-driven targeting saw a 22% increase in ROAS (Meta Engineering). The system genuinely rewards good inputs. The catch is that most accounts turning it on haven't done the diagnostic work to know whether their inputs qualify. They enabled the automation hoping it would supply what their funnel was missing, and automation doesn't work that way. It multiplies what's there. If what's there is strong, the multiplication is worth a lot. If it's weak, you've just automated the weakness.

The Underlying Fix

If your Advantage+ is underperforming, the job is to find which input is weakest. Two of them you check inside your own account: whether weekly conversion volume is high enough for the system to learn, and whether your event data is clean. But the inputs most accounts actually get wrong are the ones on the creative side — whether your hook, copy, and offer communicate something specific enough for Andromeda to match well, and whether your landing page honors the promise the ad made. Those are the hardest to judge on your own work, because you built all of it and you're too close to see the seams.

That's the work FunnelBoost is built to do. It scores your hook, copy, offer, message alignment, and landing page, and tells you which of the five is dragging delivery down — the exact creative and funnel signals Advantage+ optimizes against. $99, results in under an hour — start your audit at funnelboost.io.

Q: Why is Meta Advantage+ performing worse than my manual campaigns?

A: Advantage+ amplifies whatever signals you feed it, so if your creative is vague, your funnel is misaligned, or your conversion data is thin, it scales those weaknesses instead of your performance. Manual campaigns can mask weak inputs with tight manual targeting; Advantage+ exposes them. The fix is usually improving the inputs, not reverting to manual.

Q: Should I turn off Advantage+ if it's not working?

A: Not before checking the inputs. Turning it off treats the symptom, not the cause. Run through creative signal clarity, landing page alignment, event match quality, weekly purchase volume, and audience overlap first. In most cases one of those is the real problem, and switching back to manual just hides it again rather than fixing it.

Q: How much conversion volume does Advantage+ need to work?

A: As a directional benchmark, roughly 30–50 purchases per week gives the system enough signal to model your buyer and optimize well; below about 20–30, it often can't exit the learning phase at all. Volume requirements vary by account, but thin conversion data is one of the most common reasons Advantage+ underperforms — the model can only learn from events it receives.

Q: Does Meta Advantage+ work for small budgets?

A: The constraint is conversion volume more than budget size. A small budget that still generates consistent weekly purchases can work; a budget too small to produce enough conversion events won't give the system anything to learn from. If your spend can't generate ~20–30+ purchases a week, manual campaigns or a different objective may serve you better until volume grows.

Q: What's the difference between Advantage+ and manual Meta campaigns?

A: Manual campaigns put targeting, placement, and delivery decisions in your hands. Advantage+ hands those decisions to Meta's automation — Andromeda and GEM — which reads your creative and conversion data to make them for you. Advantage+ can outperform when your creative, funnel, and data are strong; manual can outperform when you have specific knowledge of a narrow buyer that the creative signals don't yet capture.

Related reading: What Is Meta Andromeda? · How to Audit Your Meta Ads for the Andromeda Era · The Complete Meta Ads Audit Guide · FunnelBoost home