How to Regain Targeting Control Inside Advantage+ and…

The short answer: You can't reclaim the granular audience dials Meta took away — but you can reassert intent signals through creative architecture, input constraints, and campaign structure choices that guide the algorithm without fighting it. Here's how.

The Control You Lost Is Not Coming Back (And That's the Wrong Frame)

If you've been running Meta ads for more than two years, you remember the feeling of precision. Layered interest stacks. Lookalikes built on your best buyers. Placement-level bid controls. You knew who you were talking to, and you could prove it in the data.

Then came Andromeda — Meta's infrastructure overhaul that fundamentally changed how the auction and delivery system works — and a lot of that precision evaporated. Long-running campaigns that had been profitable for years fell off a cliff. ROAS became inconsistent. Audience targeting controls were either removed outright or made functionally irrelevant by the algorithm's tendency to expand beyond them anyway.

The instinct for experienced media buyers is to fight this. To rebuild the old structures. To add more exclusions, tighten the audience, and force the machine back into a box it no longer fits in.

That instinct is costing you money.

The frame that actually works isn't "how do I get my controls back." It's "how do I feed the algorithm better inputs so it makes better decisions on my behalf." That's a different problem — and it has real, practical solutions.

Why Advantage+ Feels Like a Black Box (And Why It Partially Is)

Meta's Advantage+ Shopping Campaigns (ASC) and Advantage+ Audience are designed to give the algorithm maximum latitude. Meta's own documentation describes Advantage+ Audience as a "suggestion" to the system, not a hard constraint. The algorithm will honor it when it agrees with you and override it when it doesn't.

That's not a bug. It's the product working as intended — optimizing for Meta's definition of a conversion, which may or may not align with your definition of a valuable customer.

This is where the frustration is legitimate: the platform's optimization objective is not perfectly aligned with your business outcome. Meta wants to serve the impression that is most likely to generate a reported conversion event. You want to acquire customers who have high lifetime value, low return rates, and fit your brand. Those are related goals, but they are not the same goal.

Understanding that gap is the first step to working around it intelligently.

What Advantage+ Actually Optimizes For

Reported conversion events — whatever pixel event you're optimizing toward, weighted by Meta's modeled probability
Delivery efficiency — reaching people at the lowest cost per reported event, which can mean prioritizing easy converters over high-value new customers
Engagement signals — creative that generates interaction gets more distribution, regardless of whether that interaction correlates with purchase intent

None of these are bad objectives. But if your pixel is firing on low-quality purchases, or your creative is driving curiosity clicks rather than buyer intent, the algorithm will scale the wrong behavior — efficiently and at volume.

The Levers You Still Control (And How to Use Them)

Here's the practical reality: you have more influence over automated campaigns than the platform's UI suggests. The levers have just moved. They're no longer in the audience panel — they're in your creative, your signal quality, and your campaign architecture.

Lever 1: Creative as Targeting

This is the most important shift in post-Andromeda media buying, and it's the one most performance marketers trained on audience segmentation find hardest to internalize. Your creative is now your primary targeting mechanism.

When you run broad or Advantage+ campaigns, the algorithm uses creative engagement signals to self-select the audience. An ad that speaks directly to a specific problem, uses vocabulary your ideal customer uses, and addresses their specific objection profile will naturally attract that person — and the algorithm will learn from those engagement patterns and find more like them.

Generic creative doesn't just underperform creatively. It actively degrades your targeting by sending the algorithm mixed signals about who your customer is.

Practical implications:

Write ad copy that repels the wrong buyer as clearly as it attracts the right one. If you sell premium B2B software, your creative should feel expensive and specific — not broad and accessible.
Use customer language, not category language. "Stop losing deals because your proposal looks like a template" targets better than "The best proposal software for sales teams."
Vary creative by awareness stage, not by audience segment. Prospecting creative should assume zero context. Retargeting creative (even in broad campaigns) can reference specific objections or proof points.

Lever 2: Signal Quality Over Signal Volume

The algorithm is only as smart as the data you feed it. If your pixel is tracking every micro-conversion — add-to-cart, initiate checkout, view content — and you're optimizing on purchase, the system has a rich signal set to work from. If your pixel is poorly implemented, firing duplicate events, or missing mobile conversions, you're asking the algorithm to navigate with a broken compass.

Before you restructure a single campaign, audit your signal quality:

Is your Conversions API (CAPI) implemented and deduplicating correctly against pixel events?
Are your purchase events firing on actual confirmed purchases, not order confirmation page loads that include returns?
Have you passed customer value data (order value, LTV) back to Meta so it can optimize for high-value conversions rather than just conversion count?
Is your EMD (email, phone, name) match rate above 40%? Below that, your custom audiences and lookalikes are working from a thin foundation.

Value-based optimization — passing actual revenue data back to Meta — is one of the most underused levers available to advertisers right now. It directly tells the algorithm that a $500 purchase is worth more than a $30 purchase, which is information it cannot infer on its own.

Lever 3: Campaign Architecture as a Guardrail

You can't lock the algorithm into a specific audience, but you can use campaign structure to create meaningful separation between budget pools and optimization objectives.

A structure that works well in the current environment:

Campaign 1 — Advantage+ Shopping (Prospecting): Broad, creative-led, optimizing on purchase value. This is where you let the algorithm range freely. Feed it 6–10 creative variations across at least 3 distinct concepts. Let it run for a minimum of 7–14 days before drawing conclusions.
Campaign 2 — Manual CBO (Warm Audiences): Separate budget, separate campaign. Target your existing customer list, website visitors (90–180 days), and video viewers. This is where you maintain intentional control over who sees your retention and upsell messaging. Do not merge this into your ASC.
Campaign 3 — Controlled Test Campaign: Small budget (10–15% of total), manual placements, specific creative hypotheses. This is your learning environment. It exists to generate directional data, not to scale.

This architecture gives you a clear separation of intent: the algorithm has freedom where freedom helps (cold prospecting), and you maintain deliberate control where it matters most (warm audiences, retention, high-value segments).

Lever 4: Audience Inputs as Signals, Not Constraints

Inside Advantage+ Audience, you can still provide an "audience suggestion" — interests, demographics, behaviors. The algorithm will use this as a starting point and expand from there. Most advertisers either ignore this input entirely (leaving the algorithm with no starting context) or treat it like the old interest targeting (expecting it to act as a hard constraint).

The right approach is to treat it as a directional signal. Put in your best-performing interest clusters and demographic ranges — not because the algorithm will stay there, but because it gives the system a meaningful starting point for its initial delivery decisions. As it accumulates conversion data, it will self-correct. Your input just shortens the learning curve.

Similarly, use your existing customer list as an "Advantage+ audience" input even if you're running prospecting. The algorithm uses this to find lookalikes without you explicitly creating a lookalike audience. It's a softer version of the old 1% LAL, and it still works.

Lever 5: Spend Efficiency Controls That Actually Hold

Broad automation does not mean you have to accept unconstrained spend behavior. A few controls that remain meaningful:

Bid caps vs. cost caps: Cost caps tell the algorithm to stay at or below your target CPA on average — it will still make some bids above the cap. Bid caps are harder constraints that limit individual auction bids. In volatile delivery environments, bid caps can prevent the algorithm from chasing expensive inventory during peak hours, though they can also throttle delivery if set too aggressively.
Dayparting (ad scheduling): Still available at the ad set level in manual campaigns. If your conversion data shows that purchases cluster between 6pm–11pm in your target timezone, scheduling delivery to those windows reduces wasted impressions — though it also constrains the algorithm's ability to find efficiency across the full day. Test before committing.
Placement controls: Advantage+ placements are Meta's default and generally outperform manual placement selection at scale. But if your product category has documented performance disparities by placement (e.g., Reels dramatically underperforms for your creative format), manual placement selection is still available in non-ASC campaigns and worth testing.
Spending limits at the campaign level: Daily and lifetime budget caps at the campaign level are hard constraints the algorithm cannot override. Use them to prevent runaway spend during algorithm learning phases.

The Creative Volume Problem: How Much Is Enough?

One of the most common structural errors in post-Andromeda campaigns is running too few creative variations. The algorithm needs variety to find efficiency — if you're running two or three ads in a broad campaign, you're asking it to optimize with almost no options.

A practical framework for creative volume by budget tier:

Under $5K/month: 4–6 ads across 2–3 distinct creative concepts. Focus on concept diversity (different hooks, different angles) rather than format diversity. One strong concept in two formats beats six variations of the same concept.
$5K–$25K/month: 8–12 ads across 4–6 concepts. Introduce format diversity here — static, video, carousel. Test one variable at a time where possible, but don't let testing discipline prevent you from feeding the algorithm enough variety to work with.
$25K+/month: 15–25 ads, systematic concept testing, dedicated creative refresh cadence. At this budget level, creative fatigue is a real and measurable problem. Plan for 2–3 new concept introductions per month, not just new executions of existing concepts.

Creative fatigue is real and accelerating. The signal that a concept is fatiguing isn't just declining CTR — it's increasing frequency among your warm audiences and declining thumbstop rate among cold ones. Watch both metrics, not just ROAS, which lags behind creative fatigue by 1–2 weeks.

What "Working With the Algorithm" Actually Looks Like in Practice

To make this concrete: here's how a media buyer who has internalized these principles approaches a new campaign build differently from someone still trying to reclaim the old controls.

Old approach (fighting the algorithm):

Tight interest stacks, 3–5 layers deep
Strict age and gender restrictions
Placement exclusions based on historical preference
Single ad set per audience segment
Optimizing on lowest-funnel event available
Pausing and restarting based on daily performance swings

New approach (directing the algorithm):

Broad or Advantage+ audience with a directional interest input as a starting signal
Creative differentiated by awareness stage and buyer persona — the creative does the audience selection work
Value-based optimization with order value passed back via CAPI
6–10 creative variations per campaign, 3+ distinct concepts
Separate campaigns for prospecting vs. warm audiences — never merged
Evaluation windows of 7–14 days minimum before structural changes
Budget decisions made on 7-day rolling averages, not daily fluctuations

The second approach doesn't feel like control in the old sense. It feels like setting good conditions and then being patient — which is genuinely uncomfortable if your expertise was built on active, granular management. But it's what the current system rewards.

When to Push Back on Automation (And When Not To)

Not every campaign belongs in Advantage+. Part of regaining control is knowing when to use the automated systems and when to deliberately step outside them.

Advantage+ works best when:

You have a clear, trackable conversion event with sufficient volume (50+ purchases per week at minimum)
Your product has broad appeal and the algorithm genuinely has a large pool to find buyers in
Your creative is strong enough to self-select the right audience
You can tolerate a learning phase without pulling the plug early

Manual campaigns still make sense when:

Your conversion volume is too low to feed the algorithm meaningful data (under 20–30 purchases/week)
Your product has a very specific, narrow buyer profile that the algorithm consistently misses
You're running brand campaigns where reach precision matters more than conversion efficiency
You're in a regulated category where audience restrictions are legally required, not just preferred

The honest answer is that for most DTC advertisers spending over $10K/month with a functional pixel and sufficient conversion volume, Advantage+ campaigns outperform manual campaigns over a 30–60 day window — not because the algorithm is smarter than you, but because it has access to more real-time signals than any human buyer can process. The job is to make sure those signals are high quality and that your creative is doing the audience selection work the algorithm can't do on its own.

Reporting in a Broad Automation World: What to Actually Measure

One of the most corrosive effects of the shift to broad automation is what it does to reporting. When you can't attribute performance to specific audiences, the temptation is to fall back on platform-reported metrics — ROAS, CPM, CTR — that look clean but may not reflect real business outcomes.

Platform-reported ROAS is increasingly unreliable as a standalone metric. Attribution windows, view-through conversions, and modeled conversions all inflate the number. What you need alongside it:

MER (Marketing Efficiency Ratio): Total revenue divided by total ad spend, across all channels. This is the north star metric that doesn't lie. If your platform ROAS is 4x but your MER is 1.8x, the platform is taking credit for organic and email conversions.
New customer acquisition rate: What percentage of your reported conversions are actually new customers? Advantage+ has a documented tendency to over-index on existing customers and warm audiences because they convert more easily. If you're not tracking new vs. returning customer splits, you may be paying to re-acquire people who would have bought anyway.
CAC by cohort: What did it cost to acquire customers in this campaign, and what is their 30/60/90-day revenue? This is the only way to know if the algorithm's efficiency is real or just front-loaded.
Incrementality testing: At least quarterly, run a holdout test — turn off a campaign for a segment of your audience and measure the revenue impact. This is the only way to know how much of your attributed revenue is actually incremental. Meta's Conversion Lift tool can do this, as can third-party tools like Northbeam or Triple Whale.

Honest reporting is not just a nice-to-have. It's the mechanism by which you know whether your "working with the algorithm" approach is actually working — or whether you're just getting better at rationalizing a system that isn't delivering for your business.

The Human Role in an Automated System

The media buyers who are thriving in the post-Andromeda environment are not the ones who found a way to restore granular audience controls. They're the ones who redirected their expertise toward the inputs the algorithm can't generate on its own.

The algorithm can optimize delivery. It cannot:

Understand why your customer buys — the emotional and situational context that makes a creative concept land
Identify which creative concepts are worth testing based on customer insight, not just historical performance data
Make strategic decisions about budget allocation across channels based on business context
Recognize when the data is misleading because of attribution problems, seasonality, or external factors
Know when to stop scaling and protect margin

These are the skills that matter now. Not because human judgment is inherently superior to algorithmic optimization, but because they're the inputs the algorithm depends on to do its job well. The best media buyers in this environment are the ones who think of themselves as algorithm directors rather than algorithm operators.

Frequently Asked Questions

Can I still use interest targeting inside Advantage+ campaigns?

Yes, but it functions as a directional suggestion, not a hard constraint. Inside Advantage+ Audience, you can provide interest and demographic inputs that the algorithm uses as a starting point. It will expand beyond those inputs as it gathers conversion data. Treat your audience input as a signal to shorten the learning curve, not as a fence that keeps delivery contained.

How many creatives do I need to run in an Advantage+ campaign?

A minimum of 6 ads across at least 3 distinct creative concepts is a reasonable floor for most budgets. The algorithm needs variety to find efficiency — running 2–3 ads gives it almost no room to optimize. At higher budgets ($25K+/month), aim for 15–25 ads with a systematic refresh cadence to combat creative fatigue.

Should I merge prospecting and retargeting into one Advantage+ campaign?

No. Keep them in separate campaigns with separate budgets. Advantage+ Shopping Campaigns will naturally over-index on warm audiences because they convert more easily — if you merge prospecting and retargeting budgets, you'll often find the algorithm spending the majority of your budget re-engaging existing customers rather than acquiring new ones. Separate campaigns give you visibility and control over that split.

What's the minimum conversion volume needed for Advantage+ to work effectively?

Meta's own guidance suggests 50 optimization events per ad set per week as a target for exiting the learning phase. In practice, campaigns with fewer than 20–30 purchases per week often struggle to give the algorithm enough signal to optimize meaningfully. If your volume is below that threshold, consider optimizing on a higher-funnel event (add to cart, initiate checkout) and using manual campaigns with tighter structure until your volume grows.

Is platform-reported ROAS still a reliable metric?

Not as a standalone metric, no. Platform ROAS is inflated by view-through attribution, modeled conversions, and credit for purchases that would have happened organically. Use it as a directional signal, but pair it with Marketing Efficiency Ratio (total revenue ÷ total ad spend), new customer acquisition rate, and periodic incrementality testing to get an honest picture of what your campaigns are actually driving.

Does Advantage+ work for every business type?

No. Advantage+ performs best for businesses with broad product appeal, sufficient conversion volume, and a functional pixel setup. Businesses with very narrow buyer profiles, low conversion volume, regulated audience requirements, or products that require significant education before purchase often see better results with manual campaign structures. The honest answer is that channel-fit assessment should come before campaign structure decisions — not every business should be running Advantage+ as their primary approach.

How do I know if my campaigns are suffering from creative fatigue vs. audience burnout?

Creative fatigue typically shows up first as a declining thumbstop or hook rate among cold audiences — people are seeing the ad but not stopping to engage. Audience burnout shows up as rising frequency among warm audiences alongside declining conversion rates. If frequency is rising and cold audience engagement is still healthy, you likely have an audience saturation problem. If cold engagement is declining too, it's a creative problem. Watch both metrics together rather than relying on ROAS alone, which lags behind both signals by 1–2 weeks.

At Ise AI, we work with media buyers and growth teams who are done being sold black-box automation and want to actually understand what's driving performance. If you're navigating post-Andromeda Meta strategy and want a second set of eyes on your campaign architecture, get in touch.

How to Regain Targeting Control Inside Advantage+ and Broad Automation — Without Fighting the Algorithm