Advantage+ vs manual campaigns: what you give up and how to compare them fairly

Meta's developer blog presents Advantage+ shopping campaigns as automated ecommerce advertising that uses machine learning for audience, creative, budget, optimisation and destination¹. A later post on the same blog is titled "Advantage+ campaign experience for sales and app"². A software vendor's guide says Meta renamed the shopping format to Advantage+ sales campaigns³. Our glossary entry on Advantage+ sales campaigns covers the definition.

In plain terms, the machine decides most of what a media buyer used to decide: who sees the ad, which creative runs, how budget is spread and what the delivery system optimises for. You supply budget, creative and a goal. Meta does the rest. A related piece is Meta Andromeda, the system that shortlists candidate ads before ranking.

That is a fair deal if the machine is better than you at the rest. It is a poor one if you cannot see what it did.

Less than the marketing implies, and more than the dashboard shows. Four things matter in practice.

• Audience. You steer with signals, not with hand-built segments. You cannot easily tell which audience produced which sale.
• Placements. Delivery can shift to wherever the system finds cheap results. Your placement-level reading gets thinner.
• Creative reading. Automation tests combinations for you. That helps, but it blurs which idea worked. It also makes creative fatigue harder to spot before results slip.
• Who you reach. Jon Loomer, a Meta ads educator, has criticised Advantage+ shopping campaigns because you could not exclude people who had already bought the product being promoted⁴. If the format spends on existing customers, the platform's ROAS looks fine and your new-customer volume does not.

A 2023 trade-press headline put it as "results or control"⁵. Our view: that framing still holds. You get one, and you pay for the other.

Our view: the loss that costs money is the last one. Many existing customers would have bought anyway, so spend on them flatters the report.

Less than both sides suggest.

Meta's developer blog cites higher ROAS and lower cost per action as benefits reported by existing users, with no method, sample or baseline given¹. That is a platform describing its own product, with nothing you can rerun. It is a claim, not an independent result.

On the other side, an agency has published a two-week test for one ecommerce client in which a standard product-ad campaign beat its Advantage+ counterpart⁶. The same agency says that in its other tests Advantage+ sometimes came out ahead⁶. One test, one account, one agency with a view. It tells you manual campaigns can win. It does not tell you they will win for you.

A measurement vendor has also compared Advantage+ with manual campaigns across its own clients' experiments⁷. It still comes from a company that sells incrementality testing. Read it as one more view, not as a ruling.

So the honest summary is short. Gains reported by Meta are platform claims. Single agency tests point the other way. Neither generalises. Your account is the only sample that counts.

Most comparisons fail before they start, because the two arms are not alike. Fix that first.

1. Pick the decision. Are you choosing a default format for the next quarter, or checking one product line? Write it down.
2. Match the arms. Same products, same creative library, same country, same period. Do not run Advantage+ on your best sellers and manual on the rest.
3. Separate the audiences. Overlap between the arms muddies both. Split by geography or by product range, and avoid arms that bid on the same people.
4. Set the yardstick before launch. Use incremental ROAS (iROAS), meaning revenue the ads caused divided by ad spend, plus new-customer CAC (nCAC), which divides acquisition spend by first-time customers only. Platform ROAS is the wrong yardstick here, because it is the number Meta grades itself on. See platform reported ROAS vs incremental ROAS.
5. Allow for the Meta learning phase. Both arms need time to settle before you read anything.
6. Decide what a draw means. If results are close, which format do you keep? Agree this in advance.

Example (illustrative): you run Advantage+ on one set of regions and manual campaigns on a matched set. Advantage+ shows a platform ROAS of 5.0, manual shows 4.0. Advantage+ looks better. Then you look at new customers: Advantage+ brought 300, manual brought 450, for similar spend. Measured on nCAC, manual is cheaper. The platform said one thing. The customer count said another.

The numbers above are invented to show the logic. They are not a result from any account.

Automation does not remove your job. It moves it.

• Customer exclusions. Check which existing-customer controls the current setup gives you, and use them. Meta's interface changes, so read the live screen, not an old guide.
• Budget and scale. Raise spend in steps. Big jumps cloud the comparison.
• Creative supply. In our view, automation needs variety to work with. A thin library runs out fast.
• Conversion signal quality. The machine optimises on the events you send. Bad events mean confidently wrong optimisation.
• An independent check. Keep a measurement method that Meta does not grade, such as a holdout test or a geo split. Otherwise the platform is marking its own homework.

Our view: use Advantage+ where your volume and creative are enough to feed it, and keep a manual setup where you need to see and steer the spend. Either way, judge it on incremental results.