Google Has Put a Number on Your First-Party Data — and It's Selling the Number

Google announced a set of measurement changes on September 10 that are easy to file under "ads tooling" and skip. One of them is worth more attention than that: a Data Strength Uplift metric in Google Ads that claims to calculate the additional conversions your first-party data setup recovered.

That's a metric whose job is to justify spending on the thing being measured. Which doesn't make it wrong — it makes it worth reading carefully.

What was actually announced

Data Manager is expanding beyond Google Ads. It's being integrated directly into Google Analytics and Display & Video 360, with enhanced conversions available in both for matching customer data. The Data Manager API is also becoming universal, built on the IAB Tech Lab's Event and Conversions API standard — a genuinely useful piece of standardisation if you've ever maintained bespoke integrations per platform. Built-in diagnostics have been added to surface data issues automatically.

Data Strength Uplift arrives in Google Ads as a metric calculating conversions recovered by your first-party data configuration.

Meridian, Google's open-source marketing mix modelling platform, gets agentic capabilities for auditing data quality and guiding model building, support for brand signals such as Branded Google Query Volume, and backend performance work. Meridian GeoX is now generally available globally after this year's beta, for running causal geo-experiments across advertising platforms.

The numbers, and how to hold them

Google's announcement carries several performance claims: Data Manager users see an average 26% incremental ROAS increase; enhanced conversions produce an 11% increase in Search conversions versus standard imports; advertisers using the Google tag gateway "observe on average a 14% conversion uplift," rising above 20% for Demand Gen campaigns.

These are Google's figures about Google's products, drawn from advertisers who chose to adopt them. That last part is the load-bearing one. Advertisers who implement enhanced conversions and a tag gateway are, as a group, advertisers with the engineering capacity and measurement maturity to implement things properly. Attributing their performance to the feature rather than to the kind of advertiser who adopts the feature is a selection effect, and averages of this sort rarely control for it.

Take them as directionally plausible and as a statement of what Google wants credited. Don't take them into a forecast.

The interesting bit is Meridian, not the metric

The most substantive item here is Meridian GeoX reaching global general availability, and it's the one getting the least attention.

Geo-experiments are one of the few genuinely causal measurement tools available to marketers — you hold out regions, you change spend, you measure the difference against a counterfactual. That's a real experiment, not an attribution model reallocating credit after the fact. In an environment where signal loss has made deterministic attribution steadily less trustworthy, the methods that don't depend on tracking individuals are the ones that hold up.

That Meridian is open source matters too. You can inspect the model. You cannot inspect Data Strength Uplift, which is a number Google calculates about the value of Google's own recommendations, using a methodology Google hasn't published.

Adding Branded Google Query Volume as a brand signal is a sensible modelling improvement and also a neat piece of positioning — it makes Google Search demand an input to your measurement of upper-funnel spend like TV and out-of-home.

What to do with this

Treat Data Strength Uplift as a directional indicator, not evidence. It's fine for spotting that something changed. It is not the number to put in a board paper arguing for first-party data investment, because the entity computing it has an interest in the answer.

Look at the ECAPI standardisation properly. If you maintain multiple server-side conversion integrations, an API built on a common IAB standard is the change here with real engineering value, and it's the least promoted item in the announcement.

Take the diagnostics seriously. Automated surfacing of data issues in Data Manager is unglamorous and probably the most immediately useful thing on the list. Broken conversion imports tend to be discovered by accident, weeks late.

If you have budget scale, look at GeoX. Geo-experiments are hard to run well and hard to argue with once run. If you're spending enough for holdouts to be statistically meaningful, a causal answer beats a modelled one — and beats a vendor's metric about the value of its own advice.

Sources

• Google: New updates to measurement suite in Google Ads
• Google: Google Marketing Live 2026 — growth in the age of AI
• Google Meridian (open-source marketing mix modelling)