Meta Ads Strategy 2026: What Actually Works After Advantage+ Ate Manual

Meta ads strategy in 2026 starts from one fact: Andromeda, the new ads retrieval system, reads your creative before deciding who sees it. CPMs are up roughly 20% year over year, Advantage+ Sales now drives most ecommerce spend on the platform, and brands testing 20+ new creatives a month are reporting 65% higher ROAS than the ones still shipping under ten. Targeting moved into the ad.

That single shift, more than any policy change or attribution window cut, is what separates the accounts pulling away from the ones quietly bleeding margin. The targeting screen still exists. It just doesn't matter the way it used to.

I've been watching the post-Advantage+ scramble for about a year now. Everyone agrees the rules changed. Almost nobody agrees on what the new rules actually are. So this is the working playbook, with the receipts.

Andromeda quietly rewrote the targeting question

Meta deployed Andromeda across Instagram and Facebook in late 2024 as a next-generation ads retrieval engine. Meta's own engineering team reported a 6% improvement in retrieval recall and an 8% lift in ads quality on selected segments after rollout. Those numbers sound modest in isolation. They aren't. They're the difference between Meta showing your ad to a "probably interested" user and the user it actually thinks will convert.

Here's what it means in practice: Andromeda evaluates the creative first, then predicts engagement, then bids. Your audience setup is a hint, not an instruction. Search Engine Land's coverage of the Andromeda+GEM stack describes it as a two-stage system where retrieval and ranking are now modeled together with creative features as a primary signal.

So the old playbook (build airtight personas, stack 50 ad sets, prune the losers) doesn't get worse, exactly. It just gets quietly outperformed by accounts shipping more creative variation against broader audiences. The teams winning Meta in 2026 are the ones who treat creative production like a media buying activity, not a brand activity.

The takeaway most strategists I talk to underrate: if Meta is reading the ad to find the audience, the ad has to communicate the audience. Lifestyle setting, on-camera person, copy register, opening frame. All of those are now targeting signals, not aesthetic choices.

Advantage+ Sales is the default. Here's what its real numbers look like

Meta's claim is that Advantage+ Shopping (now branded Advantage+ Sales) cuts cost per purchase by an average of 12% versus business-as-usual campaigns. Meta documents the campaign experience in its business help center, and most third-party reports land in a similar range with wide variance. Lunio's audit of the campaign type notes that real-world ROAS tests have come in at 3.90 for ASC versus 2.39 for the equivalent manual setup, though that comparison favors a clean account with diverse creative and a working CAPI feed.

That last part matters. Advantage+ doesn't fix a broken account. It exposes one. The conditions that make ASC actually outperform are unforgiving:

• 50 conversion events per week minimum to reliably exit the learning phase. Below that, performance is unstable and CPAs run 20-50% above post-learning averages.
• $100/day floor for the optimization to compound, with $150-300/day as the practical starting point for small-to-medium ecom.
• 15-50+ active creatives rotating in the campaign. Meta's algorithm needs the variation to find a winner.
• Clean Pixel + CAPI with event match quality above 6.0 (7.0+ is where the system actually trusts the signal).

If you can't hit those, ASC is going to feel like a worse manual campaign that you have less control over. From what I've seen, that's where most of the "Advantage+ doesn't work for us" complaints come from. The campaign type is fine. The account isn't ready.

The other thing nobody warned advertisers about: Advantage+ campaigns will happily eat your retargeting budget if you let them. The "existing customer budget cap" control inside ASC is the lever that decides whether new acquisition or repeat purchases dominate the spend. The default skews toward whatever's easier to convert, which usually means existing customers. Audit that cap weekly, especially if your blended ROAS is suddenly looking suspiciously good.

The creative volume problem nobody planned for

This is the part that breaks teams. Andromeda's retrieval-first model only works if there's something to retrieve. So Meta now needs more variation than the average in-house creative team can produce.

The reported benchmark from the agencies I trust most: brands shipping 20+ net-new creatives per month are seeing about 65% higher ROAS than those producing fewer than ten. That's not a soft trend. That's a structural advantage that compounds every month a competitor lags.

And the format math has shifted too. Roughly 90% of Meta's ad inventory is now vertical, and 98% of users are on mobile. UGC-style creative (raw, vertical, social-native) is outperforming polished brand creative by 48% on CTR and 26% on CPA across most verticals. We covered the Andromeda creative fatigue curve in detail: the typical UGC ad now has a 14-18 day useful lifespan before it needs to be retired or recut.

Do the math on that. If you need 15-50 active creatives at any time and each one lasts about two weeks, you need to be producing somewhere between 30 and 100 new assets a month just to maintain a stable volume. That's a creative ops problem, not a campaign management problem.

The teams I see solving this aren't hiring more in-house designers. They're standing up a UGC pipeline with three creators on retainer, scripting in batches, and rotating hooks every Monday. The cost lands somewhere between a junior creative hire and a freelance illustrator. The output is 4x what a single in-house designer can produce.

One thing worth flagging: Meta's new Creative Workflow now reports per asset instead of per ad set, which makes the volume strategy much more measurable than it was even nine months ago. If you're still bucketing creative performance at the ad set level, you're flying blind on the part of the funnel that matters most.

Attribution is broken in a new way. CAPI is no longer optional

Meta removed the 7-day view and 28-day view attribution windows in January 2026, then redefined what counts as a click in March. Combined with the iOS 14.5 fallout that's been compounding for years, attribution accuracy on Meta has degraded somewhere in the 40-60% range, depending on the vertical. DOJO AI has the cleanest summary of what changed and where the new gaps sit.

Conversions API isn't a nice-to-have anymore. Triple Whale's setup walkthrough on CAPI is the most practical reference I've seen. The implementation matters more than the existence of CAPI: Meta's own conversion lift studies show advertisers who add CAPI on top of the pixel see a 13-19% increase in attributed conversions, and accounts running CAPI + Pixel together with proper event deduplication can recover 20-30% of lost conversion data.

The single highest-impact tweak: pass hashed email addresses with every event. That move alone can raise your event match quality score by up to 4 points. If your EMQ is sitting at 5 and you push it to 8, you'll often see Andromeda's bidding model start trusting you within a week, which usually shows up as a quiet 10-15% drop in CPA without any campaign changes.

This part is unglamorous and most agencies skip it because the client doesn't see the work. That's why the accounts that get it right open up a sustained edge.

Where manual still beats Advantage+ (and where it doesn't)

Manual campaigns aren't dead. They're just narrower in scope. The cases where I'd still reach for an old-school ABO setup over ASC:

• New product launches with no conversion history. ASC needs signal. A brand-new SKU has none. A manual campaign with a small interest stack and a $50/day budget will get you to first conversions faster.
• B2B and high-consideration verticals. If your buyer pool is genuinely small (think enterprise SaaS, accountants, dentists in a single state), Advantage+'s broad targeting is going to waste impressions. Manual with a tight lookalike still wins.
• Pure creative testing. Advantage+ rotates creative invisibly and reports unevenly. If you actually need to know which hook beats which, run a manual A/B with two ad sets and a controlled budget.
• Geographic precision. ASC's location targeting is country-level by design. Local services businesses still need manual.

The framework most agencies have settled on: 70-80% of Meta budget into Advantage+ Sales as the primary acquisition engine, 20-30% into manual ABO for retargeting, creative testing, and any audience the broad model can't find. That split tracks roughly with what I've seen working in 2026, with the caveat that "70-80%" creeps up over time as the account matures.

One nuance worth holding onto: don't run manual and ASC against the same audience at the same time. Meta's bidding logic gets confused by the overlap and will deprioritize one or the other in ways that aren't transparent. Either separate them by funnel stage (manual for retargeting, ASC for prospecting) or run them in different time windows.

The counterargument: when handing the wheel to Meta is still wrong

I want to take this seriously because most coverage of Advantage+ won't.

The honest argument against full ASC adoption is that you're voluntarily reducing your understanding of your own account. Meta's reporting inside ASC tells you what spent and what converted. It doesn't tell you which audience segment, which placement, or which combination of signals actually drove the result. You lose the ability to learn things about your customer that aren't already encoded in your creative.

For a brand still figuring out who its audience is, that loss is real. You're trading short-term efficiency for long-term knowledge. Some of the smartest paid social leads I know still run a manual structure on 30-40% of their budget specifically because they want to keep learning, even if it's a few percent more expensive.

There's also the bidding-stack concern. When most of your spend goes through ASC and ASC is making decisions you can't audit, you're trusting Meta's incentives are aligned with yours. They mostly are. They aren't always. Lunio's audit of 2.7B clicks found Google and Meta both ship a non-trivial percentage of low-quality traffic that gets billed at the same rate as the good stuff. ASC's automation makes that harder to detect, not easier.

And the third thing: incrementality. Last-click attribution has been a polite fiction for years, and Meta's reported numbers inside ASC are the most last-click-flavored of all. The accounts that genuinely know what's incremental run separate GeoLift tests or holdout experiments quarterly. That's a few thousand dollars and a couple weeks of patience. It's the cheapest way to know whether your reported ROAS is real.

So the strongest version of the contrarian case is: don't go all-in on Advantage+ if you still need to learn things about your customer, and don't trust ASC's reported numbers without an incrementality check at least twice a year. That's a defensible position. I just don't think it's the same as "don't use Advantage+."

The 2026 Meta ads stack that survives the next algo update

Pulling all of this together, here's the structure I'd build into a fresh ecom account today, knowing what we know about Andromeda and where the platform is heading:

1. One Advantage+ Sales campaign as the primary acquisition vehicle. 70-80% of the budget. Minimum $150/day. Existing customer cap set deliberately, audited weekly.
2. One manual ABO campaign for retargeting against website visitors and engaged audiences. 15-20% of budget. This is also where you run any specific creative test you need controlled data on.
3. One always-on creative testing protocol that produces 20-40 net-new assets per month, with a Monday rotation. Track per-asset performance, not per-ad-set.
4. CAPI + Pixel with EMQ > 7.0, hashed email on every event, deduplicated. This is the single non-negotiable. Without it, the rest of the stack underperforms by 15-25% silently.
5. Quarterly GeoLift or holdout test to validate incrementality. Budget the cost into your annual measurement plan, not into a campaign line item.
6. A creative ops function (could be one person and three contractors) that's measured on volume, hook diversity, and per-asset CPA, not on brand consistency or polish.

That's the stack. None of it is exotic. The hard part is committing to all six pieces at once, because skipping any one of them tends to break the assumptions the others rely on.

If you're wondering where this fits with the rest of your paid budget, our 70/20/10 framework for Google Ads budget allocation uses a similar logic for the same reason: leave room in the structure for the platform's automation to work, but reserve a slice for the things only a human can decide.

Meta ads strategy in 2026: the questions everyone keeps asking

Is Advantage+ Sales the same as Advantage+ Shopping?

Yes. Meta renamed it in late 2024. The campaign type is functionally the same, with some updated controls (most notably the existing customer budget cap and the new total store impact attribution view added in March 2026). If you've seen old documentation referring to ASC, it's the same product.

What's the minimum budget for Advantage+ Sales to actually work?

$100/day is the floor where Meta's algorithm starts accumulating useful signal. $150-300/day is where most small-to-medium ecom accounts see ASC outperform their old manual setups. Below $100/day, you're likely better off with a single manual campaign and tighter targeting until you've built up conversion volume.

Should I still use detailed targeting in 2026?

Mostly no, with exceptions. Detailed interest targeting is still useful for new accounts with no conversion history, very small budgets, or genuinely niche markets where the buyer pool is identifiable. For most accounts, a 1-3% lookalike combined with broad targeting and strong creative will outperform any interest stack. The targeting moved into the ad.

How many creatives do I really need to be running on Meta?

The benchmark from accounts hitting 65% higher ROAS is 20+ net-new creatives per month, with 15-50 active in any given campaign at once. UGC-style ads have a 14-18 day useful lifespan, so the volume isn't optional. If you can't sustain that, prioritize fewer creatives with more iteration cycles (different hooks, same body) over fewer creatives held longer.

Is CAPI worth the engineering work in 2026?

Yes, and it's not really optional anymore. Meta's own conversion lift studies show CAPI on top of the pixel adds 13-19% in attributed conversions on average. With proper event deduplication, total conversion data recovery sits in the 20-30% range. The single biggest implementation lever is passing hashed emails on every event, which can lift your event match quality score by up to 4 points and meaningfully change how Meta's bidding model treats your account.

How do I know if Meta's reported ROAS is actually real?

Run a GeoLift test or a holdout experiment at least twice a year. Pause Meta in 10 matched DMAs for four to six weeks while keeping it on in 10 control DMAs, then measure the actual sales delta. The result will almost certainly differ from what Meta reports inside the platform. The gap between the two numbers is what's known as incrementality, and it's the only honest measure of paid social performance.

What about Advantage+ Audience versus Advantage+ Sales?

Different products. Advantage+ Audience is a targeting layer you can apply within almost any campaign type to let Meta find your buyers. Advantage+ Sales is the full automated campaign type for ecom. You can use Advantage+ Audience inside a manual campaign, but you can't really opt out of it inside an Advantage+ Sales campaign, since the automated targeting is the whole point.

Where I'd actually start tomorrow

If you're inheriting a Meta account that's been running on the old playbook, the order I'd attack it in is roughly: fix the CAPI implementation first (biggest leverage, hardest to see), audit the existing customer cap on any Advantage+ campaigns second, build out the creative volume pipeline third, and only then start moving budget around between campaign types. The temptation is always to restructure the campaigns first because it feels like the most visible change. In my experience that's the move that produces the least durable improvement.

Meta's not going to stop iterating on Andromeda or the campaign types. By next year there will be another rebrand, another set of automation features, probably another attribution window cut. The accounts that stay ahead aren't the ones chasing every update. They're the ones with creative volume, clean signal, and a measurement plan that doesn't depend on Meta telling them the truth.

Honestly, that's the whole strategy. Everything else is implementation detail.