In March 2026, thousands of advertisers watched their conversions fall off a cliff. CPAs jumped. ROAS sank. Nothing about the campaigns had changed.
Meta had changed how it counts.
That single event explains why attribution settings stopped being a background configuration and became a thing media buyers argue about in client calls. If you do not know which of five layers your account is running, you are making budget decisions on numbers that do not mean what you think they mean.
Here are the Meta ads attribution settings best practices that hold up in 2026.
What Attribution Settings Actually Control
Attribution settings decide two things at once, and most advertisers only think about the first.
Reporting. Which conversions appear in Ads Manager and get credited to which ads.
Optimisation. How the algorithm learns. The window defines what counts as success, so it shapes delivery as well as the report.
That dual role is why changing a window is never just a reporting decision. Narrow the window and you also narrow what the algorithm chases.
The Five Layers in Meta’s 2026 Attribution Stack
Set up a website conversion campaign today and you are running five distinct settings, whether you chose them or not.
| Layer | Default | What It Does |
|---|---|---|
| Click-through window | 7 days | Credits conversions after a link click |
| Engage-through window | 1 day | Credits conversions after a social interaction or qualifying video view |
| View-through window | 1 day | Credits conversions after an impression with no interaction |
| Attribution model | Standard | Counts everything in the window, or Incremental predicts causation |
| Conversion count | All conversions | Counts every event, or only the first per person |
Each layer affects both what gets reported and how delivery is optimised. Miss one and you misread the account.
What Changed in March 2026
Four changes landed together, and the effects are still working through most reporting.
Click-through narrowed to link clicks only. Previously, likes, shares, saves, comments, and carousel swipes all counted as clicks and triggered the full 7-day window. Now only genuine link clicks do.
Engage-through arrived as a new category. Everything that left the click bucket landed here: social interactions and qualifying video views. It is fixed at one day and cannot be extended.
The video threshold dropped from ten seconds to five. The old bar was calibrated for long Feed video and made little sense for Reels, where much of the purchase impact traces to the opening seconds.
The 28-day view-through window was retired as a reporting option.
Meta’s stated goal was closer alignment with third-party analytics, which have always counted only real link clicks.
Why Your Conversions Dropped and Performance Did Not
Two things happened to those reclassified conversions, and only one of them is neutral.
Some moved buckets. A like followed by a purchase the next day now appears under engage-through instead of click-through. The conversion still exists, just in a different column.
Others vanished. A share on Monday followed by a purchase on Friday used to earn full 7-day click credit. Now that share buys a 24-hour window, so the Friday purchase is attributed to nothing at all. The interaction still influenced the sale. Meta no longer has anywhere to record it.
The practical implications:
- Do not compare pre-March and post-March data directly. You are comparing two measurement systems.
- Set a fresh baseline from mid-March 2026 and evaluate forward from there.
- Check your engage-through columns before concluding a campaign broke.
- If you apply multipliers to Ads Manager numbers based on incrementality or a third-party source, update those factors.
There is no true legacy view. Historical time series will not mirror the old model cleanly.
Choosing the Right Window by Business Type
There is no universal correct setting. There is a right starting point for your purchase cycle.
| Business Type | Suggested Starting Window | Reasoning |
|---|---|---|
| Impulse DTC, low AOV | 1-day click, or 7-day click | Short consideration, fast decisions |
| Standard ecommerce | 7-day click, 1-day view | Enough signal to optimise, captures realistic timelines |
| High AOV or considered purchase | 7-day click | Longer research, but view-through inflates heavily |
| Lead generation | 1-day click | Decision is immediate; longer windows add noise |
| B2B with 2–4 week cycles | 7-day click, validated externally | Expect meaningful under-reporting inside the window |
That last row deserves attention. B2B advertisers with multi-week sales cycles can lose a substantial share of attributed conversions to the 7-day ceiling. Those conversions still happen. Meta simply stops counting them, which is an argument for CRM-based validation rather than a longer window.
The View-Through Question
The default includes 1-day view, meaning someone who saw your ad without clicking and purchased within 24 hours counts as a conversion.
For brands with heavy organic traffic, large email lists, or aggressive retargeting, this over-attributes. Those people were often going to buy anyway.
The diagnostic is straightforward. Pull the breakdown by click versus view in Ads Manager. If view accounts for more than roughly a quarter of your reported conversions, you have a judgement to make about how much of your reported ROAS is real incremental lift.
Many practitioners now recommend 7-day click only as the cleaner reporting lens for DTC, while leaving the broader window in place for optimisation. Others keep the default for delivery quality and simply report both. Both positions are defensible. Running view-through without knowing its contribution is not.
Standard vs Incremental Attribution
Meta’s incremental attribution option, rolled out during 2025, uses modelling and holdout testing to count only conversions it predicts would not have happened without the ad.
Standard attribution counts everything in the window regardless of whether the person would have purchased anyway. Incremental attempts to answer the question standard attribution ignores.
Two honest cautions.
Selecting incremental removes your ability to edit attribution settings. You trade control for a different measurement philosophy.
And the numbers can be startling. One widely circulated vendor case reported prospecting ROAS falling from around 8x under standard attribution to under 2x under incremental. Treat that as a single documented example rather than a benchmark, but treat the direction as real.
Incremental is most useful for high-value conversions and for brands already running holdout tests. It is a poor first move for an account that has not yet fixed its conversion tracking.
Rules for Changing Attribution Settings
Attribution changes register as performance changes. Handle them accordingly.
- Never change mid-flight on a performing campaign. It resets the optimisation signal and the trendline becomes meaningless.
- Change at natural breakpoints. New campaign launch, creative refresh, budget restructure.
- Watch the learning threshold. Moving from 7-day to 1-day click can cut attributed conversions substantially, potentially dropping an ad set below the roughly 50 weekly conversions it needs to optimise reliably.
- Document every change as an explicit breakpoint in your reporting, with a dated annotation.
- Never compare accounts or periods running different windows. This is the single most common cause of a wrong conclusion in a client review.
How to Validate Meta’s Numbers
Platform-reported figures carry the platform’s own assumptions. Validate against sources that do not.
Use Compare Attribution Settings. Ads Manager lets you view the same campaign across different windows without changing live settings. The gap between 7-day click only and the full default tells you how much view-through is lifting your reported numbers.
Cross-reference GA4. Different tools, different models. GA4 defaults to data-driven attribution with a long lookback, while Meta runs its own stack. Document the gap and use it as a standing benchmark rather than trying to make the numbers match.
Track blended metrics. Marketing efficiency ratio and blended customer acquisition cost do not care about attribution windows. If Meta reports improving ROAS while blended CAC worsens, believe the blended number.
Run holdout tests. Geo holdouts and conversion lift studies answer the incrementality question directly.
Reconcile against back-end revenue. Your payment processor is the only true source.
Reporting Practices That Prevent Arguments
- Separate click-through and engage-through columns rather than reporting a combined total.
- State the attribution window on every report, every time.
- Present platform-reported and blended figures side by side.
- Annotate March 2026 as a permanent methodology break in any historical chart.
- Agree a single reporting window with the client and stick to it for a full quarter.
Presenting total conversions without the breakdown makes it impossible to judge the quality of each attribution type, and guarantees confusion the moment a stakeholder opens GA4.
Mistakes That Cost Real Money
- Leaving the default without knowing what it reports.
- Reacting to the March 2026 reclassification by cutting spend on campaigns that never actually declined.
- Changing windows mid-campaign, then reading the resulting noise as performance.
- Comparing this quarter to last quarter across the methodology break.
- Running pixel only, without Conversions API, so the whole attribution question rests on incomplete data.
- Treating incremental attribution as a fix for broken tracking.
A 30-Minute Attribution Audit
Do this today.
Minutes 1–10. Open Ads Manager, check the attribution setting at ad set level, and confirm whether engage-through is on. Note the five layers as currently configured.
Minutes 11–20. Run Compare Attribution Settings for the last 30 days. Record the gap between 7-day click only and the full default. That gap is your view-through inflation.
Minutes 21–30. Compare Meta’s reported ROAS against blended MER for the same period. Document the ratio and make it your standing benchmark.
You now know how much of your reported performance is click-driven, how much is inferred, and how far Meta’s view sits from your business reality. Every budget decision after that is better informed than it was this morning.
FAQs
What is Meta’s default attribution setting in 2026?
Seven-day click, one-day engage-through, and one-day view-through, using the standard attribution model and counting all conversions.
What is engage-through attribution?
A category introduced in March 2026 covering conversions after social interactions or video views of at least five seconds, fixed at a one-day window.
Why did my Meta conversions drop in March 2026?
Reclassification, not performance. Non-link clicks moved from the seven-day click window to a one-day engage-through window, and some conversions fell out entirely.
Should I use 7-day click or 7-day click plus 1-day view?
Use the broader window for optimisation signal and the click-only view for cleaner reporting. Check what share of conversions view-through contributes first.
Is incremental attribution worth using?
It suits high-value conversions and brands already running holdout tests, but it removes your ability to edit attribution settings and will report lower numbers.
