Q2 2026 brought one of the most confusing periods Meta advertisers have faced in years. Many apps saw conversions drop, took it as a signal that something broke, and scaled their budgets back. In many cases, it wasn’t necessary.
Before changing anything—budgets, campaigns, creative—it helps to understand what actually happened. There are two distinct forces behind the recent drop, and they require very different responses.
To help separate the two and figure out what to actually do, we asked Samet Durgun, an app growth consultant and fractional CMO who has navigated this exact shift across multiple accounts, to walk through what he’s seeing and what’s worked.
What’s actually happening
1. Meta’s March 2026 attribution update
Meta now only counts actual link clicks as conversions. Likes, saves, and video views have been removed from the conversion path.
Earlier, in January, Meta had already removed its 7-day and 28-day view-through attribution windows. These two changes together led to significant reported conversion drops across many accounts—even where actual sales hadn’t moved. Click-through windows are a separate setting and weren’t affected.
A few concrete steps help adjust to this shift:
- Confirm engage-through attribution is enabled in Ads Manager (it’s on by default with a 1-day window).
This is a new metric introduced in March 2026 that captures non-click interactions (video views, likes, saves, shares) that convert within a day.
Turning it on recovers the engagement-based conversions that disappeared from click-through.
- Adjust the click-through attribution window to match your actual conversion timeline. For subscription apps with longer consideration periods, 7-day click is now the practical default.
- Stop comparing this month’s reported conversions directly to pre-March numbers — they’re no longer measuring the same thing, so month-over-month comparisons need an asterisk.
Knowing where to look for reliable data matters as much as knowing what changed. Samet Durgun, mobile growth specialist, points to a source most teams underuse:
App Store Connect is the source of truth. People really need to pay attention to their downloads and where they come from. Go to Analytics, First Time Downloads, click “By Source Type,” and you’ll see that App Referrer will be your click-attributed installs from Meta and other apps. Search could include view-through influenced users who saw your ad but searched the App Store later. And App Store Browse is usually featuring or ranking high in popular keywords. If you’re not checking this breakdown regularly, you’re guessing which installs are whose.
2. The Andromeda algorithm shift
Meta’s new AI-powered delivery system moved from optimizing for the metric you set (clicks, conversions) to predicting the long-term value of a user. To do this accurately, it needs more and cleaner conversion signal than before. Creative became the primary targeting signal: Meta now reads the ad first and finds the audience for it, rather than the other way around.
As a result, audience fatigue appears faster, narrow interest targeting is less effective, and creative libraries filled with visually similar ads often perform as though only a single ad is running.
The instinctive response to all of this — editing campaigns, turning things on and off, testing new audiences — tends to make things worse, not better. Every structural change resets Meta’s learning phase. Samet describes the pattern clearly:
The biggest mistake is panic-editing. Most teams blame the ads first. It’s often wrong. Making frequent daily changes resets Meta’s learning phase. If something looks off on day two, resist the urge to change everything. Go through the checklist first: did you change creatives? Campaign settings? Is your campaign getting enough events? Was there an SDK change? A product or paywall change? Or was it just a strong or weak cohort?
How to spot it
Outbound clicks remain stable, but reported conversions drop sharply after March 2026. CRM data, orders, and revenue tell a very different story from Ads Manager. If that’s the case, the gap is likely a measurement issue rather than a performance issue.
What to do
When the numbers don’t add up, the question is where to look first. Samet explains his process:
Your App Store Connect and your subscription management service are your source of truth. They show what actually happened. Cross-reference your MMP and Meta numbers against them. If those numbers disagree by more than 20%, start checking your event setup, your CAPI configuration, and whether your MMP is deduplicating correctly. If business metrics — revenue, LTV, retention — are stable but Ads Manager numbers dropped, that’s a measurement issue.
Once the gap is identified, the practical steps are:
- Avoid restructuring campaigns in response to reporting changes alone. Pausing campaigns, rebuilding ad sets, or aggressively changing budgets can create real performance problems where none existed before.
- Compare Meta-reported conversions with CRM and revenue data before and after the March attribution update.
- Shift performance reporting toward business outcomes: revenue, LTV, and subscriber growth rather than platform-reported conversions.
When Meta’s reporting becomes unreliable as a day-to-day signal, there is a simpler metric that keeps teams grounded. Samet recommends:
Blended ROAS. Total revenue divided by total ad spend. It includes revenue from organic users Meta didn’t acquire, conversions from view-based attribution, and differences in attribution windows between sources. Attributed ROAS and blended ROAS tell very different stories. You need both, but when Meta’s reporting breaks, blended is what keeps you grounded.
Problem 2: Meta is seeing only a fraction of your conversions
How to spot it
- Event Match Quality is below 6.5/10.
- iOS conversion rates are significantly lower than Android.
- With iOS 26 (September 2025), Apple started stripping Meta’s click ID from all Safari sessions by default — meaning even users who clicked your ad may arrive on your page with no attribution attached.
- Events Manager shows browser-side events only, without corresponding server-side events.
What to do
The core fix is giving Meta the signals it needs to optimize effectively. This means moving beyond pixel-only tracking toward a proper server-side setup. Samet describes what a trustworthy tracking setup looks like in practice:
If possible for your app, server-side CAPI with hashed email, phone, name, and external ID alongside the client-side SDK. Meta scores how well it can match your events using an Event Match Quality score from 0 to 10. Client-side only gets you a 3–5. Add server-side and you can reach 8 or higher. That’s the difference between Meta matching maybe half your events versus 80–90% of them. You want at least a 6 on purchase events.
For subscription apps specifically, the practical steps are:
- Implement Conversion API alongside the pixel and configure event deduplication using event_id.
- You already have the data Meta needs: emails (and phone numbers in some cases) from web onboarding and IP address from the server. Make sure both are passed with every conversion event server-side.
- Ensure that trial starts, purchases, and renewals are all tracked server-side.
Problem 3: Creative fatigue is happening faster than you think
How to spot it
- Frequency exceeds 3.0.
- CTR is declining from baseline.
- Ads have been running for more than a week without refreshes.
- Active creatives look visually similar despite different copy variations. Under Andromeda, these creatives may effectively be treated as the same ad.
What to do
The most common mistake is treating creative diversity as a visual question—different colors, different thumbnails, different layouts. Under Andromeda, the algorithm reads creative far more deeply than that. Samet explains:
Different scripts, not different thumbnails. If you lead with “save time,” you attract busy people. If you lead with “save money,” you attract price-sensitive users. Each USP pulls a different audience. Meta looks at who engages with your ad and finds more people like them. Your creative shapes your audience through what it says and through who responds to it. If all your ads say the same thing with different visuals, Meta might treat them as one ad.
The practical actions:
- Refresh creatives every 7–10 days based on frequency and CTR signals rather than a fixed schedule.
- Audit your creative library for visual and messaging similarity. Different copy alone is no longer enough.
- Build a continuous creative production process with weekly testing targets.
- Use broader targeting and allow creative variation to do more of the audience selection work.
- Treat funnel testing as an extension of creative testing. The funnel continues the story started by the ad.
Check these before changing your campaigns
- Check your verification status in Business Manager (Authorizations and Verifications)—even if you’ve verified before, as it has lapsed for many accounts. This is an unofficial but circulating recommendation from Meta’s internal team. Unverified accounts see silent delivery restrictions that look like performance drops but have nothing to do with campaigns.
- Also verify your identity directly at facebook.com/ID — select Confirm Identity, choose the first option, and submit ID for review.
- Review Account Quality for policy warnings or content restrictions.
- Check whether CPM inflation is market-wide before assuming it’s a campaign problem. If competitors in your vertical are seeing the same increases, it’s structural—not something campaign changes will fix.
- Use Audience Overlap to identify ad sets competing against each other.
- Audit the post-click experience before changing campaign settings, including load speed, checkout flow, pricing changes, and paywall performance.
Staying ahead of the next drop
The accounts navigating 2026 well share an operating model more than a set of tactics. Four things stand out:
- Creative production is treated as ongoing infrastructure with defined output goals and refresh schedules.
- Tracking is maintained proactively through regular pixel, CAPI, and Event Match Quality reviews.
- Revenue metrics serve as the primary source of truth—not Ads Manager reported conversions.
- Early warning indicators such as rising frequency, CPM inflation, and declining CTR are monitored weekly.
The takeaway
Most of what looked like a collapse in Q2 2026 was three things layered on top of each other: a change in how Meta counts conversions, a shift in how its algorithm decides who sees your ads, and the usual instinct to react fast when the dashboard turns red. Taken apart, each of these has a clear diagnosis and a clear fix. Taken together, they’re easy to misread as one big problem—and that’s when teams end up making changes that cost them more than the original drop did.
The accounts that came through this in good shape didn’t do anything dramatic. They checked their data sources, gave Meta better signal, kept their creative moving, and gave the algorithm time to adjust before judging the results.
