If your Meta campaigns suddenly looked worse in early March 2026, you weren't alone. Meta rolled out a significant update to its AI delivery system that changed how the algorithm optimizes for conversions, and advertisers felt the ripple effects quickly.
The core change: Meta shifted from auction-based placement optimization to outcome-based optimization. Instead of optimizing toward clicks and placement signals, the algorithm now tries to predict downstream conversions, the actions that actually matter to the business. In principle, that's a good thing. In practice, the transition triggered a wave of reported CPM increases of 15% to 40%, along with performance degradation in campaigns that had been optimized for clicks or landing page views.
Before you restructure your entire account in response, it's worth diagnosing whether what you're seeing is a genuine performance problem or a reporting one.
This is the first question to answer, and it requires looking beyond Meta's platform attribution.
Under the new model, Meta is taking credit for fewer conversions. That's by design. If someone saw your ad, liked or shared but didn't click the link, searched for your brand 3 days later, and converted through Google, Meta no longer counts that. Previously, view-through and assisted attribution inflated platform numbers, thereby flattering Meta's reported ROAS. Outcome-based optimization produces a cleaner, more conservative view of what Meta is actually driving.
So if your Meta platform numbers dropped but your blended performance, revenue, pipeline, and back-end CRM data held relatively steady, the update may have changed how conversions are attributed more than it changed how your campaigns are actually performing.
Breaking down your conversion data by attribution window will tell you a lot here. Compare 1-day click, 7-day click, and view-through attribution separately. If view-through volume collapsed but click-based conversions stayed consistent, your campaigns are likely healthier than the dashboard suggests. Meta platform numbers should now more closely align with what you see in cross-platform measurement tools and back-end data, which, for many advertisers, will feel like a drop even when the underlying business result hasn't changed.
The practical implication: back-end data and blended measurement are now the only way to accurately evaluate Meta performance under this model.
The reported CPM increases of 15% to 40% across retail, lead generation, and ecommerce are not a glitch. Meta's algorithm is now pricing higher-intent impressions more aggressively because it believes they're more likely to lead to actual conversions. You're paying more per impression because the impressions are, in theory, better qualified.
That said, "in theory" is doing real work in that sentence, and advertisers still need to manage costs on a few fronts.
Audience scope is one lever. Smaller, tightly defined audiences tend to carry higher CPMs; there are fewer impressions available, and competition for them is intense. Widening your targeting can reduce costs, but there's a real tradeoff: broader audiences may include lower-intent users whose conversion behavior dilutes downstream performance. A lower CPM that comes with a meaningfully lower CVR may not be a win. Test incrementally rather than opening the aperture all at once.
Creative is the other lever and arguably the more important one. Under outcome-based optimization, creative quality and freshness have become the primary performance drivers, surpassing audience-targeting parameters, which the algorithm now handles more autonomously. Meta's AI surfaces ads to receptive audiences on its own; what it can't do is make weak or stale creative perform.
Test a wider variety of ad types and angles. Images tend to carry lower CPMs than video in general; that's a pattern worth factoring into your mix. But beyond format, increasing the cadence at which you're introducing new creative angles is the main defense against fatigue-driven CPM escalation. New creative gives the algorithm more material to work with and creates opportunities to find lower-cost pockets before they get crowded.
The new AI system is more data-hungry than the old one. Campaigns generating fewer than 50 weekly optimization events lost meaningful algorithmic priority in the update; the system simply doesn't have enough signal to optimize effectively at that volume.
If you're running multiple campaigns that each produce modest conversion volume, consolidation is worth serious consideration. Concentrating conversion signal into fewer ad sets gives the algorithm what it needs to find and prioritize high-intent users. Fragmented campaign structures that spread volume too thin are now being penalized more explicitly than before.
For accounts genuinely constrained by conversion volume and teetering on the edge of the learning phase exit with low weekly event counts, it may be worth optimizing for a higher-funnel action that still indicates quality. A completed lead form or a key page engagement can serve as a more achievable optimization signal than a purchase or qualified lead submission, provided the proxy event is actually predictive of downstream value. This requires knowing your funnel well enough to make that call confidently.
We recommend making two fairly substantial strategic mindset adjustments based on the new world of outcome-based Meta optimization. The first is creative testing velocity; the second focuses on how you evaluate Meta’s back-end reporting.
For creative, the update makes it important to ramp up the cadence and variety of testing options. Companies with limited creative resources might consider outsourcing creator content or testing AI creative generation tools to get a higher volume and more variety to feed into an accelerated testing schedule.
For reporting, the shift to outcome-based optimization reinforces what should already be true about how you evaluate Meta: platform numbers are not the source of truth for business impact.
That recommendation has been on the table for years, largely due to limitations in tracking loss and attribution. The March update makes it more urgent. Advertisers who rely on Meta's reported conversions as their primary performance signal will consistently misread what's happening, both when numbers look artificially strong and when they look artificially weak.
The right frame is understanding how Meta activity influences your overall bottom line, not how many conversions the platform claims. That means investing in cross-platform measurement, monitoring back-end data alongside platform metrics, and being willing to hold the tension between what Meta reports and what your business is actually seeing.
Meta will keep making updates. Some will help, some will create disruption, and most will require adaptation. The accounts that stay nimble, testing creative consistently, reading blended data, not over-indexing on any single platform signal, are the ones best positioned to absorb changes like this without overreacting in either direction.