Meta Andromeda Update: How to Keep Your App‑Ads Performing 

You have probably heard about the big update from Meta around July 2025. For a lot of advertisers the cost‑per‑install (CPI) or click‑to‑install (CTI) has increased, even though you did what “worked before”. It’s no coincidence. The Meta Andromeda update has shifted the rules of ads, and if you’re running ads for an app (especially a small to mid‑sized growth team), this matters.

In this article, we’ll dive into what exactly Andromeda is, why it’s so relevant for performance marketers, and most importantly: how you adapt your campaigns so you’re not left behind. We’ll walk through:

• the tech behind the change
• how targeting, creatives & campaign structure must evolve
• the direct KPI implications (CPI, ROAS, install‑to‑purchase, LTV)
• a step‑by‑step action plan

Actionable real‑world insight from what we’ve seen in the field. 

‍

What is the Meta  Andromeda update?

The technical basis: retrieval engine & AI at scale

Meta describes Andromeda as an “innovative end‑to‑end hardware, software, machine‑learning co‑designed system … introduced in 2024 … enabled a 10,000× increase in the complexity of models used for ads retrieval.”

What that means in practice: instead of the ad‑delivery system looking at a moderate pool of ads and then deciding which to show to which audience, Andromeda first filters millions of possible ads and narrows them down to a few thousand candidates per user scenario. 

When and how did the rollout happen (timeline)

Though the engineering blog released in December 2024, many advertisers reported noticing material shifts around mid‑2025 (July/August) when the system gained full reach. 

For growth teams this means: if your performance started changing around that time, it’s likely tied to this update, not just seasonal variation or competitor budget.

Why Meta made the change (ad‑retrieval bottleneck & AI flood)

Why now? Because the old system was dying. With the rise of Advantage+ campaigns, more and more ad creatives, generative AI, and creative turnover, Meta’s retrieval layer hit a bottleneck. In simple terms: too many ads, too many combinations, and the old engine couldn’t efficiently decide which should even get shown. 

For advertisers: this means your old “tight audience + one hero creative” playbook is less effective now. The engine expects input diversity. It expects you to feed it variety so it can do its job.

Unique insight: For app‑growth teams, this is an opportunity, not a threat. If you treat creative variation and messaging as your targeting trick (rather than obsessing over narrow audiences), you can benefit from this update rather than be disrupted by it.

‍

Why this update matters for advertisers

The old way vs the new way of Meta/Instagram ads

If you’ve been running ads for apps the past 2‑3 years, you likely followed a pattern: define a handful of specific audience segments (lookalikes, interests), test maybe 3‑6 creatives per ad set, find the “winner”, scale. That model still works sometimes, but under Andromeda, this might not perform anymore.

In the new version:

• The algorithm cares less about how tightly you define your audience, and more about how well your creative can match the person.
• The “who” (audience) becomes less precise, and the “what” (creative message/format) becomes more varied.

‍

Creatives become the new targeting mechanism

As one analyst put it: Your creative is your audience. The idea: Instead of you segmenting the audience and hoping the right message hits, you provide multiple different messages (for different personas/desires/awareness levels) and the algorithm picks which user sees which message.

For an app this might mean: instead of having “users aged 25‑34, interest = fitness” and one ad “Download now”, you build multiple creatives: one targeting busy parents who want quick workouts, one for college students wanting social fitness, one for fitness enthusiasts wanting performance upgrades. The message isn’t just the same with different copy, it’s genuinely different angle + format.

Broad vs tight targeting: what changes

One of the biggest shifts: you may want to shift from hyper‑segmented ad sets to broad targeting + creative variation. Meta’s internal guidance (and publishing) suggest that when creative variation is strong, broader targeting works better under Andromeda. 

For SMBs & app growth teams, this means fewer ad sets, more creatives, and ensuring your budget isn’t fragmented across dozens of micro‑audiences.

‍

What it means for KPI’s: CPI, ROAS, install‑to‑purchase, LTV

Because the algorithm is changing how it matches ads to users, you’ll likely see movement in your core metrics:

• CPI (Cost Per Install) may temporarily go up if you continue with the old model and the algorithm doesn’t favour your creatives.
• CTI (Click to Install Rate) may drop if your creatives aren’t resonating as cleanly with audiences.
• But if you adapt, you can improve install‑to‑purchase conversion, sessions per user, LTV, and ultimately ROAS because you’re reaching better‑fit users.
  Real‑world: Some teams we know saw CPI jump by ~20‑30% initially until they adapted. They then regained and even improved their ROAS by focusing on creative variation and quality rather than just volume.

For app businesses, this shift is critical. Not just acquiring users, but acquiring valuable users (who stick, convert, and generate LTV). Andromeda gives the machine more options to find them, if you feed it.

‍

The impact on campaigns for app‑businesses and SMBs

Cost increases (CPM, CPC, CPL) & why you’re seeing them

Many ad‑accounts reported suddenly higher CPCs, CPMs, and CPS (cost per sale) around July/August 2025. This aligns with the full rollout of Andromeda. When your creatives don’t provide the algorithm enough variety or don’t resonate strongly, you lose efficiency.

What causes the cost bump:

• Insufficient creative diversity → algorithm can’t match well → fewer impressions, higher cost.
• Fragmented ad structure → limited data per creative → slower learning → higher budget needed.
• Old‑method creatives get quickly flagged as “not differentiated” → less spend.

Retention, sessions, LTV affected by targeting/creative change

Let’s talk about what happens after the install. For apps, the install is only part of the story. If the new targeting/creative set brings lower‑quality users (e.g., users who install but don’t engage), you’ll see retention drop, uninstall rate rise, sessions per user decline, and LTV suffer.

Under Andromeda your creative is influencing which persona you attract and thus what behaviour they show. A well‑crafted message for a high‑intent user will more likely result in higher session count, quicker onboarding, and higher ARPU. So adaptation here has knock‑on effects beyond installation metrics.

‍

New rules: creative diversification & campaign structure

Why “more of the same” doesn’t cut it anymore

A key message across multiple sources: minor tweaks (a different hook, different CTA) won’t be enough under Andromeda. The system is better at pattern‑recognition and will treat visually similar creatives as the same.  In other words: if your “new ad” just has the same background, same actor, same set‑up, slightly changed words, the algorithm may treat it as duplicate and the spend will go to other more varied creatives. You lose differentiation.

Framework for creative diversity (Persona‑Desire‑Awareness or P.D.A.)

One effective framework: P.D.A. = Persona + Desire + Awareness level. 

• Persona: Who is the user? (Busy parent, mobile gamer, student, etc.)
• Desire: What do they want? (More playtime, fitness, connection, challenge)
• Awareness: Where are they in the journey? (Unaware of problem, problem‑aware, solution‑aware)

By combining these you build truly different ads, not small variants. For example:

• Persona: “New mom” + Desire: “Get 15 min fitness in day” + Awareness: “Problem‑aware” → Ad video: “How I got back my energy between school runs.”
• Persona: “College gamer” + Desire: “Top the leaderboard” + Awareness: “Solution‑aware” → Static image ad: “Win medals in your first week using [App]”.

This level of differentiation gives Andromeda more to match with more users.

‍

Recommended formats: video, static, carousel, length variations

Variation isn’t just in message but also medium. Sources suggest including short vertical video (6‑10s), longer video (30‑60s), static image, carousel set, slides/text overlay. 

For an app, you might find best results with:

• 6s vertical “Download now – free‑trial starts”
• 15s horizontal “See how [persona] cut time wasted on X with our app”
• Carousel “3 key features you didn’t know”
• Static “Limited time offer: 30 % off subscription”

‍

Structure of campaign/ad set: 1 campaign, 1 ad set, 10‑50 ads (broad)

According to multiple analyses: the old structure (many ad sets each narrow) is replaced by a simpler structure. One campaign per objective, one broad ad set, large creative library. For instance: “One campaign → one ad set → 10‑50 ads” with broad targeting. 

For app SMBs this means:

• Use campaign budget optimisation (CBO) where possible.
• Avoid dozens of micro‑audience ad sets unless you have a huge budget.
• Prioritise spend into fewer ad sets but richer creative variation.
• Use smart exclusions (existing users, low‑value geos) rather than tight inclusion targeting.

‍

Step‑by‑step action plan for your growth team

Step 1: Audit your existing campaigns under Andromeda

Start with facts. Analyse key metrics: CPI, CTI, install‑to‑purchase conversion, LTV (30/60 days if possible), uninstall rate. Identify ad‑sets with very low spend (“starved creatives” means algorithm didn’t favour them). Ask yourself: “How many truly different creatives do we have in this ad‑set?” If the answer is “3 versions of the same video,” you’re risking the update’s new rules.

Step 2: Creative roadmap setup (how many, what variation)

Build a creative matrix. Example:

• Month 1: Produce 15 distinct ads (5 personas × 3 desires) in mixed formats.
• Month 2: Add 20 refreshes + 5 new personas.
• Month 3: Add 30 more.
• Keep track in a spreadsheet: creative concept, persona, format, hook, date live, impressions, CPI, install‑to‑purchase. Make a budget for the creative production part of your media‑plan. Without fresh creatives you might struggle and not reach full potential.

Step 3: Adjust campaign model (targeting, structure, Advantage+)

Switch to broad targeting. Enable Advantage+ (Meta’s automation) where possible. Use one campaign per objective (e.g., “App install – Free trial” + “In‑app purchase event”). Use one ad‑set with 10‑50 creatives. Use exclusions (previous users, low‑LTV). Set value rules if needed (reduce spend on age range historically low value).

Step 4: Measure & optimise signals (tracking, events, install‑to‑purchase)

This is non-negotiable if you want to make Andromeda work for you: your SKD tracking setup must be set up correctly. The new algorithm doesn’t just guess who should see your ads, it learns from your event data. If that data is weak or incomplete (think: bots, non-qualified leads, incomplete installs), your campaign optimization performance will go down.

Here’s what you need firing accurately and consistently:

• Install event
• First_session or onboarding_complete
• Purchase / subscription / trial-start
• Uninstall or churn proxy (optional but powerful)

This is exactly where FalconMetrics comes in. We designed FalconMetrics to help small growth teams see exactly which ad + event combinations drive your real ROI, not just CPI.

With FalconMetrics, you: 

• Track full user journey: from click to install to registration or purchase
• Spot which creatives bring in high-LTV converting users 
• Built for small-medium sized teams: startup-friendly pricing, no long commitments, 30-day free trial, and 60‑day money-back guarantee if it’s not helping.

Step 5: Scale & refresh cycle (how often to refresh creatives, when to pause)

Schedule regular creative refreshes: aim for at least one meaningful refresh per month (if budget allows). At the same time, monitor creative fatigue. If frequency increases, CPC/CPI rises it’s time to pause losing creatives (bottom ~20 %) and introduce new ones. Also monitor if one creative is monopolising spend, good, but still keep variety coming so you don’t “max out” a micro‑segment.

‍

Quick Takeaways / Key Points

• The Meta Andromeda update is a major behind‑the‑scenes of Meta’s ad delivery system, creative variety is now a key driver of performance.
• For app‑marketing teams: your creative portfolio is now your targeting lever. Broader audiences + stronger creative library beats micro‑segmentation.
• Campaign structure should simplify: one campaign, one (broad) ad set, many (10‑50) creatives rather than dozens of ad sets.
• Focus not only on CPI but on quality of install, install‑to‑purchase conversion, LTV. Because Andromeda favours relevant matches.
• Avoid old mistakes: too many tiny audiences, minimal creative variation, early kills, no exclusions/value‑rules.
• Build a creative roadmap, invest in production and tracking. This isn’t a short‑term fix, it’s a strategic shift.
• Use this moment as an advantage, growth teams that adapt will pull ahead of competitors still stuck in 2023 targeting mode.

Conclusion & Call to Action

The Meta Andromeda update isn’t just another algorithm tweak. It’s a structural shift in how Meta delivers ads. From manual targeting and isolated creative testing, to machine-led matching and creative diversity at scale.

For app businesses and performance marketers, this means the old habits, tightly defined ad sets, one hero video, small tweaks, will slowly stop working. But this is the good news: this shift rewards smart and creative advertisers, not just big budgets.

If you adapt your structure, diversify your creatives, and most importantly feed the algorithm the right signals, you can outperform teams that are still operating like it’s 2023. This is no longer about big budgets. It’s about who feeds Andromeda the richest signals and most relevant creative variety.

This is exactly what FalconMetrics was built for.

If you’re a small-to-mid-sized team running install or conversion campaigns, FalconMetrics helps you:

• Track install-to-purchase conversion by creative, persona, funnel
• See which ad formats drive conversions and higher LTV
• Get going with startup-friendly pricing, free 30-day trial, and 60-day money-back guarantee

You don’t need a data science team. You don’t need an expensive agency. You need good attribution, visibility, speed, and ad winners.

Start your free trial with FalconMetrics

‍