What changed in Google Ads measurement and which numbers need checking
Published: · Sergey Zhadov
On September 10, 2026, Google brought three layers of advertising measurement into one update: first-party data pipelines, estimates of recovered conversions and causal experiments. The tools can help, but the reporting gets messy when every modeled conversion is treated as a proven sale.
Data Manager moves closer to analytics and media buying
Google announced direct Data Manager integrations with Google Analytics and Display and Video 360. The aim is to make offline events, app data and other first-party signals easier to connect across advertising tools without a separate pipeline for every account.
The company also announced enhanced conversions in those products and built-in data diagnostics. Before connecting anything, keep the boring checklist: where did the event come from, is its processing permitted, has the order been duplicated, and does the identifier match a real transaction? Otherwise a clean integration turns bad inputs into confidently reported garbage.
Data Strength Uplift is an estimate, not cash in the till
The new metric is intended to estimate additional conversions recovered by a first-party data setup. It can help compare a sparse signal set with a more complete one, for example after offline conversions or better matching are connected.
The mistake is to present a platform estimate as independent proof. Recovered conversions may improve optimization, but they still need reconciliation with CRM records, payments, refunds and deduplication rules. Google's published averages describe its stated samples; they do not promise the same result for your account.
Meridian tries to connect advertising with causal impact
The open-source marketing mix model Meridian is gaining agentic help for data quality checks, error resolution and model building. Brand signals, including branded query volume, can be added so upper-funnel activity is not reduced to the last click.
GeoX is now generally available globally. It supports geographic experiments that compare test and control regions after advertising changes. That gets closer to asking whether advertising caused sales, but only with comparable regions, enough time and no simultaneous promotion that breaks the comparison.
What to check in your account now
List your data sources and trace one real order from the CRM to the advertising report. Check duplicate handling, import delay, currency, time zone and refund rules. Only then examine what changed in Data Strength Uplift.
For Meridian, separate observational correlation from an experiment in advance. Record the business metric, period, test regions and possible confounders. If you cannot explain where a number came from, do not put it in a management report as an established fact.