Advantage+ can deliver a real uplift. Meta says advertisers using its Advantage+ shopping campaigns see 22% higher return on ad spend than those who don't, and plenty of agencies report that consolidated structures outperform the legacy builds they replaced.
Then why does everyone have a story about switching it on and watching results fall off a cliff?
Because most accounts turn it on and keep every habit that stops it from working, then judge it on day eight.
The short answer: Advantage+ underperforms when it is starved of budget, interrupted during learning, pointed at the wrong goal, fed weak conversion data, given too little creative variety, or boxed in by audience restrictions. Fix those six things and the platform usually stops being the problem.
Here are the six mistakes that cause almost all of it, and exactly what to do instead.
Mistake one. Running it alongside everything else
The most common setup is Advantage+ added as a new campaign while eleven legacy ad sets keep running beside it.
Now your budget is split across twelve entities, all bidding into the same auction, all learning from a fraction of the data. Advantage+ is a system that improves with volume, and you have deliberately starved it.
The accounts getting the uplift run most of their spend through it. Our rule of thumb is that the clear majority of budget should flow through the consolidated campaign before you judge the benefit. That means turning legacy campaigns off, which feels reckless and is the entire point.
Why does overlap hurt so much? Because your old ad sets and your new campaign are chasing many of the same people. You end up competing with yourself for the same impressions, and each campaign sees only a slice of the conversions that should be teaching it. The system cannot learn a pattern from data it never receives.
You do not have to flip everything in one afternoon. Move the bulk of spend across, keep one proven legacy campaign running at a reduced budget as a benchmark if it makes you sleep better, and switch off the long tail of small ad sets that were never pulling their weight anyway. Meta's broader shift toward fewer, bigger campaigns is covered in our piece on how Andromeda changed the rules, and it is the same logic here.
Mistake two. Judging it before it has learned
The system needs conversion volume to calibrate. Depending on your budget and conversion rate that is somewhere between several days and a couple of weeks.
During that period performance is genuinely unstable. This is not a bug. It is the cost of tuition, and it is paid once.
Most advertisers panic on day four, make three changes, and reset the learning. Then they do it again the following week. The campaign never gets past the expensive part, and the conclusion is that Advantage+ does not work.
Every change you make resets the clock. An account that gets adjusted twice a week is an account that pays tuition forever and never graduates.
So what counts as a change? Big budget jumps, swapping the optimisation event, editing targeting, and pausing or replacing most of the live creative all count. Before launch, agree a rule with whoever touches the account: no edits for the first two weeks unless something is genuinely broken, like a tracking failure or a spend runaway. Write the rule down. The temptation on day four is strong, and a written rule beats willpower.
When you do scale budget after learning, do it gradually. Large overnight jumps can knock a stable campaign back into instability, which puts you straight back on the expensive part of the curve.
Mistake three. Feeding it the wrong goal
This one costs the most and shows up the latest.
If the campaign optimises for form submissions, it will become extraordinarily good at finding people who submit forms. Cost per lead will drop, volume will rise, and everyone will be delighted for about ten weeks.
Then close rate collapses, because the system was never asked to find buyers.
Optimise for the deepest event you have enough volume to support. A booked appointment. A qualified lead marked as such in your CRM and pushed back. A purchase with a value attached. If volume is too low for the deepest event, use the next one up, and push real outcome data back regardless.
For ecommerce this is usually simple: optimise for purchase and pass the order value. For lead generation it takes more work. The fix is to connect your CRM so that when a lead becomes qualified, or becomes a sale, that event is sent back to Meta. Now the system can learn the difference between a person who fills in a form and a person who pays you. That single connection often matters more than anything else on this list.
Mistake four. Poor conversion signal
Advantage+ is a prediction engine. Predictions are only as good as the examples they learn from.
That means Pixel and Conversions API running together, deduplicated, with Event Match Quality above 7 on your primary event. Below that, the system is trying to find more people like a group it cannot clearly identify.
Most accounts have never checked this number. It is three clicks deep in Events Manager and it is worth more than your last four creative tests.
Raising it usually comes down to sending more customer information with each event, such as hashed email and phone number, and making sure browser and server events carry a matching event ID so they are counted once, not twice. We cover the full fix in the boring number quietly deciding your ad results.
Mistake five. Not enough genuinely different creative
Consolidated campaigns need creative variety to work, because the creative is now doing the targeting.
Fifteen to twenty distinct live concepts is the shape that performs. Distinct means different hooks, formats, presenters and angles, not different colour grades of one video.
This matters more than it used to because near identical ads tend to get clustered together and treated as roughly one option. Thirty variants of one idea is not thirty chances. Uploading more files is not the same as giving the system more to work with.
A simple way to get to fifteen genuinely different concepts:
- Different problems. Five ads, each opening on a different pain your customer has.
- Different formats. Talking head, product demo, static image, customer testimonial, screen recording.
- Different people. The founder, a staff member, a real customer.
- Different angles. Price, speed, quality, risk reversal, social proof.
Mix across those and you have real variety rather than the same idea in fifteen outfits. Keeping that pipeline full is a production problem, not a media buying problem, which is why we wrote up the factory for twenty new ads a month.
Mistake six. Over restricting the audience
The instinct is to help by narrowing. Add interests, exclude segments, layer a lookalike, tighten the age range.
Every restriction shrinks the pool the system can explore, and exploration is where the uplift comes from. It finds pockets of demand you would never have selected, because it is reading response patterns rather than demographic assumptions.
Keep geography. Keep age only where it genuinely matters, such as legal restrictions. Exclude recent converters. Then get out of the way.
If this feels like giving up control, remember where the control moved to. You still decide who the ad speaks to. You just do it with the hook and the message instead of an interest checkbox. An ad that opens by naming a specific problem attracts the people who have it.
The setup that works
- One consolidated campaign carrying the majority of spend.
- Deepest reliable conversion event, with real outcomes fed back from your sales system.
- Pixel and Conversions API together, match quality above 7.
- Fifteen to twenty genuinely distinct creatives live, refreshed weekly.
- Minimal audience restrictions beyond geography and converter exclusions.
- A change freeze during learning, then one deliberate change at a time.
If you are moving an existing account across, do it in this order:
- Fix tracking first. Check Pixel and Conversions API, deduplication and match quality before you touch campaigns. Launching on bad data wastes the learning period.
- Pick the optimisation event. Choose the deepest event you get enough of, and set up the CRM connection if you sell through leads.
- Build the creative bank. Have fifteen distinct concepts ready before launch, not three now and more later.
- Launch and consolidate. Move the majority of budget in, switch off overlapping legacy ad sets, keep a benchmark if you want one.
- Freeze for two weeks. Watch, take notes, change nothing unless something is broken.
- Then iterate slowly. One change at a time, starting with replacing the weakest creative.
A worked example
This is hypothetical, to show the shape of it.
Say you run a Brisbane physio clinic spending $4,000 a month on Meta. You have nine ad sets built on interests like running, pilates and CrossFit, all optimising for form fills, with four ads that are the same video in different crops. Leads are cheap. Few of them book.
You fix the Conversions API so booking events are sent from your practice software. You change the goal from form fill to booked appointment. Your team films fifteen different concepts: back pain at a desk, a runner's knee, a post surgery rehab story, a price explainer, and so on. You launch one Advantage+ campaign with most of the budget, keep your best old ad set at a small spend as a comparison, and switch the rest off.
Week one, cost per result jumps and your stomach drops. You leave it. Week two, it settles. By week three you can compare cost per booked appointment against the old setup, which is the only number that ever mattered. Whatever the result, you now know it honestly, because you gave it a clean test.
What to check before you blame the platform
Before you conclude Advantage+ does not work for your business, walk this list.
Is most of your spend actually running through it, or is it one campaign among eleven. Has it had a clean two weeks without changes. Is it optimising for something that makes you money rather than something that makes a dashboard look good. Is your match quality above 7. Do you have fifteen genuinely different creatives live, or fifteen versions of three.
In almost every account that reports disappointing results, at least three of those five answers are no. The platform is rarely the variable. The setup almost always is.
What to expect, honestly
Week one is uncomfortable. Week two starts to settle. By week three you should see whether it is working, and by week six you will have a stable baseline better than what you had before.
The businesses that get the 22% are not doing anything clever. They are doing the boring version correctly and then leaving it alone long enough for the system to do its job.
The businesses that do not are usually one panic away from a result they will never see.
If you would rather hand that setup to someone who has the patience and the process, that is what our advertising team at LIQID NOISE does. We fix the tracking, rebuild the structure, build the creative bank and then hold our nerve through the learning period, so your account finally gets the fair test it never had.
Frequently asked questions
How long does Meta Advantage+ take to exit the learning phase?
It depends on budget and how often your chosen event happens, but expect anywhere from several days to a couple of weeks. Performance is unstable during that time. Avoid big edits to budget, targeting, goal or creative while it learns, because each significant change can restart the process and extend the expensive period.
Should I turn off my old Meta campaigns when I start Advantage+?
Mostly, yes. Running many legacy ad sets alongside Advantage+ splits your budget and data, and your campaigns end up competing against each other. Move the majority of spend into the consolidated campaign. Keeping one proven older campaign at a small budget as a benchmark is reasonable, but the long tail should go.
What conversion event should Advantage+ optimise for?
The deepest event you get enough volume of. For online stores that is usually a purchase with the order value attached. For lead based businesses, aim for booked appointments or qualified leads sent back from your CRM. Optimising for cheap form fills trains the system to find form fillers, not customers.
How many ads should I run in an Advantage+ campaign?
Aim for fifteen to twenty genuinely distinct concepts live at once. Distinct means different hooks, formats, people and angles, not small edits of the same video. Near identical ads tend to be treated as one option, so volume only helps when the ideas are actually different. Refresh the weakest performers regularly.