TL;DR: Run both. Advantage+ Shopping wins on speed and volume. Manual wins on control and incrementality. Haus tested 640 Meta experiments through July 2025 and found 58% of brands got higher incremental ROAS from manual campaigns. Our verdict: put 70% of budget in Advantage+, hold 30% in manual, and read the result over six weeks.
The verdict: run both, and let the split settle the argument
Advantage+ Shopping is not a smarter version of a manual campaign. It is a different bet.
Advantage+ bets that Meta’s model finds buyers better than your audience list does. Manual campaigns bet that you know something Meta does not: which customers are worth acquiring, which creative belongs in front of which segment, and what a genuinely new customer costs you.
Both bets win in different accounts. A structured test is the only way to find out which one wins in yours.
| Dimension | Advantage+ Shopping | Manual campaigns |
|---|---|---|
| Build time | 20 to 30 minutes, one campaign | 3 to 6 hours across ad sets |
| Audience control | Country and account-level exclusions only | Interests, lookalikes, custom audiences, exclusions per ad set |
| Creative volume needed | 15 to 20 assets to feed rotation | 3 to 5 per ad set |
| Time to stable delivery | Fast, often inside 7 days at real budget | Slower, every ad set learns separately |
| Prospecting control | None since Meta retired the existing customer budget cap | Exclude buyers directly and hold a clean cold pool |
| Diagnostic data | Thin, no audience-level breakdown | Rich, you can read segment by segment |
| Incrementality risk | Higher, it leans toward warm demand | Lower, you pick the pool it buys from |
| Best fit | Catalogs over 50 SKUs, $30K+/month, steady creative supply | Restricted categories, small catalogs, high repeat purchase rates |
What Advantage+ Shopping actually automates
Advantage+ Shopping collapses the ad set layer. You set one campaign, one budget, one country, and one conversion event. Meta’s model then picks the audience, the placement, the creative combination, and the bid.
That is the whole product. It is not a new inventory source and it is not a new bidding model. It is the removal of your targeting decisions from the auction.
Meta also retired the existing customer budget cap, which used to let you force a prospecting-only split inside the campaign. The workaround now is a separate manual sales campaign that excludes your buyer audiences.
What you still control
- Total budget and the pace you raise it
- Every creative asset and the hooks in the first 3 seconds
- Product feed quality, titles, and pricing
- The conversion event you optimize toward
- Country, language, and account-level exclusions
- Attribution window, which quietly decides who looks like a winner
Four of those six are creative and feed decisions. That is the honest read on automation across both platforms, and it is the same pattern we describe in Meta’s Advantage+ and Andromeda ad stack.
Meta’s own numbers, and where they stop being useful
Meta has published one headline figure for years. Advertisers using Advantage+ Shopping campaigns saw a 32% increase in return on ad spend versus manual setups, per Meta testing reported by Social Media Today on February 28, 2023.
Adoption numbers back that up. On the January 29, 2025 earnings call, CFO Susan Li said Advantage+ Shopping campaign revenue had surpassed a $20 billion annual run rate, growing 70% year over year. By the July 29, 2026 call, she put Meta’s full Advantage+ end-to-end suite above $75 billion in annual revenue run rate.
Those are real numbers. They are also the wrong numbers for your decision.
A run rate measures how much advertisers spend through a product. It does not measure whether that spend produced sales they would not have gotten anyway. And the 32% ROAS lift is measured by the platform selling the ads, using its own attribution.
The incrementality problem nobody puts in the deck
Haus published the largest independent read on this to date. Its Meta Report, released July 28, 2025, covered 640 Meta incrementality experiments run since January 2024, across advertisers spending an average of $14 million a year on Meta.
The headline for Meta overall was good. Meta ads drove an average 19% lift to the brand’s primary business KPI. That is a platform doing real work.
The Advantage+ read was not good.
Advantage+ reported 2.4% higher ROAS inside Ads Manager. Measured incrementally, 58% of brands saw higher incremental ROAS on manual campaigns. Advantage+ averaged 12% lower direct-to-consumer incremental ROAS while spending 18% less per day.
The timing detail matters most for how you run a test. Advantage+ was ahead by 9% at the midpoint of the average experiment, then fell behind by the end. In the post-treatment window, manual campaigns held a 32% lift against 17% for Advantage+.
Read that twice before you kill a manual campaign in week two. A three-week test would have handed you the wrong answer.
The mechanism is not mysterious. Advantage+ optimizes toward the cheapest reported conversion. Your existing customers and warm retargeting pools are the cheapest reported conversions in almost every account. The model finds them, buys them, and books them as new performance.
The six-week testing framework
Week 0: fix measurement before you touch campaigns
Most Advantage+ tests fail here, not in the ad account.
- Conversions API live, with event deduplication verified against the browser pixel
- One conversion event for both campaigns, usually Purchase, never a mix
- Identical attribution setting on both, 7-day click and 1-day view
- A new-customer segment defined in your own data, not in Ads Manager
- Blended MER tracked daily from your store, not from Meta
- A geo holdout list picked in advance, matched on revenue and seasonality
Weeks 1 and 2: build the split
Put 70% of Meta budget in Advantage+ Shopping and 30% in a manual prospecting campaign that excludes all buyer audiences.
Use the same creative in both. Same offer, same landing pages, same event. The only variable you are testing is who chooses the audience.
Then leave the budgets alone. Meta’s learning phase needs roughly 50 optimization events in 7 days per ad set, per its Business Help Center. Every budget edit over about 20% risks resetting that.
Weeks 3 and 4: sit on your hands
This is the window where Advantage+ looks great and manual looks tired. Haus found exactly that pattern at the midpoint. Do not act on it.
Watch blended MER and new-customer revenue from your own store. Ignore campaign-level ROAS in Ads Manager entirely for these two weeks. It is the number under test, so it cannot also be the judge.
Weeks 5 and 6: read incrementality, then decide
Run a geo holdout or a Meta conversion lift study across the full account. Then apply three rules.
- If blended MER rose and holdout geos underperformed, shift toward 85/15 in favor of Advantage+.
- If blended MER was flat while Advantage+ reported strong ROAS, you bought existing demand. Move to 50/50 and rebuild manual prospecting.
- If new-customer count fell while total revenue held, that is cannibalization. Cut Advantage+ to 40% and exclude buyers harder.
Six weeks is the floor, not a preference. Haus experiments averaged 18.6 days of test plus an 8.8-day post-treatment observation window, and the ranking flipped inside that window.
The same discipline applies across every automated product, which is why we keep test design separate from platform choice in our AI in PPC playbook.
What our accounts show
Across iClick-managed Meta accounts, CPM now averages $11.54, up 38% year over year. CPC averages $0.70, up 8%. CTR sits at 0.90% and is declining as creative fatigues faster. Conversion rate averages 1.85%.
Run the arithmetic on those two middle numbers. At $0.70 per click and a 1.85% conversion rate, a purchase costs roughly $38 in clicks. Fifty conversions a week means about $1,900 a week in click cost before any campaign has a stable read.
That is the real budget floor for this test, and it is why brands under $20K a month rarely get a clean answer from Advantage+. They never leave learning.
The account results behind those benchmarks set the bar a test has to clear. Wish Rock Relaxation runs 10x ROAS at $80K+ a month with conversion rate up 62%. We Love Eyes runs 10x ROAS with blended CAC down 28% and Amazon orders up 340%. Zager Guitars went from $300K to $1.5M a month at 8.4 blended ROAS. Splendid Iris holds 6.2x on a $20K monthly budget. Mazaeus holds 8.4x. Lynk Pleasure holds 4.7x with zero account suspensions.
Those are blended account numbers, not Advantage+ numbers. That is the point. A new campaign type has to beat what a well-run account already produces, not beat zero. Across 250+ clients and $61M+ in managed spend since 2017, our portfolio average conversion goal lift is 41%.
The same cannibalization pattern shows up on Google, where PMax eats Shopping traffic in 80%+ of the accounts we audit. We break that down in Google’s AI ads stack: AI Max, Performance Max, and Demand Gen.
When manual still wins outright
Restricted categories. Lynk Pleasure and The Enhanced Male both sell into categories where a single wrong placement or creative pairing triggers a review. Advantage+ gives you no placement-level control. Manual keeps the account alive.
Small catalogs. Under about 20 SKUs, there is not enough product variance for the model to explore. You are paying for automation that has nothing to automate.
High repeat purchase rates. If more than 40% of your revenue comes from returning customers, Advantage+ will find them first and your reported ROAS will look excellent while new-customer count flatlines.
Anything that is not eCommerce. Advantage+ Shopping is a sales-objective product built on a catalog. Lead generation for law firms and SaaS needs a different structure entirely, because lead quality varies and the platform cannot see which leads sign.
Launches under $10K a month. You cannot clear 50 weekly events at $38 a purchase. Build demand manually first.
Five mistakes that ruin the test
- Different creative in each campaign. Now you are testing creative, not targeting, and you will not be able to tell them apart.
- Reading the test at day 14. Advantage+ was ahead by 9% at the midpoint in the Haus data and behind at the end.
- Leaving buyer audiences in the manual campaign. Both campaigns then chase warm demand and the comparison is meaningless.
- Judging on Ads Manager ROAS. Platform-reported ROAS is the claim you are testing. Use blended MER and a holdout.
- Editing budgets mid-test. Every change above roughly 20% risks a learning reset and costs you a week.
None of this is an argument that automation is overrated. It is an argument that automation needs a control group, which is the same conclusion we reached in our honest answer on whether AI replaces PPC agencies. Running that control group every quarter is most of what AI PPC management actually means in practice.
Find out which side your account is on
If you are running Advantage+ Shopping and cannot say what it did to your new-customer count, you do not have a result yet. You have a dashboard.
We will pull your Meta and Google accounts apart and send back a 47-point written audit in 5 business days. It covers your current Advantage+ split, buyer-audience leakage, feed health, and whether your reported ROAS survives an incrementality read. It is a PDF, not a sales call.
Related on iClick
Sources
- Susan Li, Meta Q4 2024 earnings call transcript, January 29, 2025 (Advantage+ Shopping past $20B annual run rate, growing 70% YoY)
- Susan Li, Meta Q2 2026 earnings call transcript, July 29, 2026 (Advantage+ end-to-end solutions above $75B annual revenue run rate)
- Haus, The Meta Report: Lessons from 640 Haus Incrementality Experiments, July 28, 2025
- Social Media Today, Meta’s automated targeting tools driving a 32% increase in ROAS, February 28, 2023
- Meta Business Help Center, About the learning phase (about 50 optimization events in 7 days)
- Meta for Business, Advantage+ Shopping Campaigns product page
- Jon Loomer Digital, No More Existing Customer Budget Cap
- iClick Advertising, 2026 PPC Benchmarks (iClick-managed accounts, 2024 to 2025)
- iClick Advertising, client case studies (Wish Rock Relaxation, We Love Eyes, Zager Guitars, Splendid Iris, Mazaeus, Lynk Pleasure)
Frequently asked questions
Is Advantage+ Shopping better than manual campaigns?
Neither wins in the abstract. Meta reports a 32% ROAS lift for Advantage+ over manual setups. Haus, testing 640 Meta experiments through July 2025, found 58% of brands got higher incremental ROAS from manual. Advantage+ usually wins on speed, volume, and build time. Manual usually wins on control and incrementality. Test both in your own account.
How long should an Advantage+ versus manual test run?
Six weeks minimum. Haus experiments averaged 18.6 days of testing plus an 8.8-day post-treatment window, and Advantage+ led by 9% at the midpoint before falling behind by the end. A two or three week read gives you the opposite of the true answer. Budget two weeks for learning, two weeks untouched, and two weeks for the incrementality read.
What budget does Advantage+ Shopping need to work?
Across iClick-managed Meta accounts, CPC averages $0.70 and conversion rate averages 1.85%, so a purchase costs roughly $38 in clicks. Meta’s learning phase needs about 50 optimization events in 7 days. That is close to $1,900 a week, or $8,000 a month, just to exit learning cleanly on one campaign.
Can I run Advantage+ Shopping and manual campaigns at the same time?
Yes, and you should. Meta’s own 2023 testing measured Advantage+ added alongside existing manual campaigns, not replacing them. Exclude your buyer audiences from the manual campaign so the two are not bidding for the same warm pool. Start at 70% Advantage+ and 30% manual, then move the split based on blended MER and new-customer count.
Why does Advantage+ report higher ROAS but drive less revenue?
Advantage+ optimizes toward the cheapest reported conversion. In most accounts, that means existing customers and warm retargeting pools who would have purchased anyway. Reported ROAS climbs because those conversions are cheap to claim. Incremental revenue stays flat because the demand already existed. A geo holdout or conversion lift study exposes the gap in about two weeks.
