Marketing Attribution in 2026: How to Measure What Works
Platform numbers, your own analytics and incrementality tests each answer a different question. Here is how to combine them in 2026.
Marketing attribution in 2026 is less about finding one source of truth and more about combining three types of measurement: platform attribution for daily decisions, your own analytics for cross-channel reporting, and incrementality tests or marketing mix models for budget decisions. Privacy changes and shifting platform rules over the past year have made relying on any single number risky.
This guide explains why your Meta, Google and store figures disagree, what changed in 2026, and how a growing business can set up measurement it can trust without an enterprise budget.
Why the numbers never match
Add up the conversions reported by Meta and Google and you will almost always get more sales than your store recorded. Each platform counts a sale it touched, using its own rules, and both can claim the same order. Three factors drive most of the gap.
- Different windows. Google Ads defaults to a 30-day click-through window, a 3-day engaged-view window and a 1-day view-through window. Meta’s default is 7-day click-through, 1-day engage-through and 1-day view-through.
- View-through credit. A view-through conversion means someone saw an ad, did not click, and bought later. Some of those people would have bought anyway.
- Missing signals. AppsFlyer reported in April 2025 that about 50% of iOS users consent to app tracking under Apple’s App Tracking Transparency. Platforms fill the gap with modelled conversions, which are estimates.
What changed at Meta in 2026
From 12 January 2026, Meta’s Ads Insights API stopped returning the 7-day view and 28-day view attribution windows. Reports or dashboards that used those windows lost them, and long view-through windows are no longer an option for comparison.
In March 2026, Meta also narrowed what counts as a click. Per Jon Loomer’s breakdown of the change, click-through attribution now only includes conversions after a click on an ad link. Conversions after likes, comments, shares, saves and other non-link clicks, along with engaged video views, moved into a new “engage-through” category with a default 1-day window. Many advertisers saw click-attributed ROAS drop overnight even though sales had not changed.
If you compare performance across early 2026, note these dates in your reports so a reporting change is not mistaken for a performance change.
Layer 1: platform attribution for daily optimisation
Platform numbers are still the right input for in-platform decisions: which ad, audience or keyword to scale. The bidding algorithms optimise on this data, so feed them clean signals.
- Send server-side data: Meta’s Conversions API and Google’s enhanced conversions recover conversions the browser misses.
- Deduplicate events so a purchase is not counted twice by the Pixel and the API.
- Pick one attribution setting per platform and keep it fixed. Changing windows mid-quarter makes trend lines meaningless.
Layer 2: your own analytics for cross-channel reporting
GA4, Shopify analytics or your CRM sees all channels, so each sale is credited once. Tag every paid link with consistent UTM parameters, and agree on one report the business reviews weekly.
A blended metric is a useful check here. Marketing efficiency ratio (MER), total revenue divided by total marketing spend, cannot be inflated by overlapping attribution. Our guide to ROAS, MER and break-even formulas shows how to calculate it. If platform ROAS rises but MER falls, platforms are probably claiming sales they did not cause.
A simple post-purchase survey (“How did you first hear about us?”) also helps, especially for channels that drive awareness but rarely get the last click, such as podcasts, influencers and YouTube.
Layer 3: incrementality and MMM for budget decisions
Attribution tells you who saw or clicked an ad before buying. Incrementality tells you how many of those sales the ad actually caused. That is the number you need when deciding where the next rupee or dollar goes.
Incrementality tests
An incrementality test holds back ads from a control group (a set of users or regions) and compares sales with the exposed group. These used to need large budgets. At Google Marketing Live in May 2025, Google said advertisers can now run incrementality tests with an experiment budget as low as US$5,000, using a Bayesian method that needs less data.
Meta now offers an incremental attribution setting at the ad set level, which optimises for conversions its models predict were caused by ads. Treat it as a directional signal. When Seer Interactive tested it across about US$1.05 million of spend, Meta reported 87% of conversions as incremental, while GA4 suggested 67%.
Marketing mix modelling
A marketing mix model (MMM) uses two or more years of weekly spend and sales data to estimate each channel’s contribution, including offline channels and factors like pricing and seasonality. It needs no user-level tracking, which makes it unaffected by cookie and consent changes.
Two free, open-source options exist. Google made Meridian available to all marketers on 29 January 2025; it uses Bayesian causal inference and can take incrementality test results as inputs. Meta maintains Robyn. Both need someone comfortable with Python or R, and both need enough spend variation across channels and weeks to produce sensible results.
A practical setup by business size
| Stage | What to use |
|---|---|
| Early (one or two channels) | Clean platform tracking, GA4 or store analytics, MER, post-purchase survey |
| Growing (three or more channels) | All of the above, plus one geo or holdout incrementality test per quarter on your biggest channel |
| Established (large, steady spend) | An MMM refreshed quarterly, calibrated with incrementality test results |
Whatever your stage, write down which number answers which question. Platform ROAS decides which ad to scale. Your analytics decide channel reporting. Incrementality and MMM decide the budget split. For a sense of what platform ROAS figures to expect, see our post on what a good ROAS on Facebook ads looks like.
Frequently asked questions
What is the best attribution model in 2026?
No single model is enough. Use platform attribution for in-platform optimisation, your own analytics for cross-channel reporting, and incrementality tests or marketing mix modelling for budget decisions.
Why do Meta and Google report more sales than my store?
Each platform credits any sale it touched within its attribution window, including view-through conversions, so both can claim the same order. Modelled conversions also add estimates.
What is incrementality testing?
It compares sales between a group that sees your ads and a control group that does not, to measure how many sales the ads actually caused rather than just preceded.
Do small businesses need marketing mix modelling?
Usually not at first. MMM needs about two years of weekly data and meaningful spend across several channels. Smaller businesses get more value from clean tracking, MER and occasional holdout tests.
