Not all conversions can be measured directly. When a user declines consent or data is missing, the platform uses an estimate instead of an exact record — a modeled conversion. Here's what that is and how to read it.
How a modeled conversion arises
When some conversions can't be attributed directly (e.g. the user declined cookies), the platform doesn't have to give up on them entirely. From the behavior of users it can measure, plus other signals, it statistically computes how many conversions likely occurred in the unmeasured part too. The result is an estimate — a modeled conversion.
Where you'll meet it
Most often in connection with Consent Mode v2 in advanced mode: even when a user declines consent, an anonymous cookieless ping is sent, and Google computes modeled conversions from it and from the behavior of those who consented. This gives you an estimate even where you'd otherwise have a zero. The context is covered in Consent Mode v2 in practice.
How to read it correctly
- It's an estimate, not an individual record. A modeled conversion doesn't correspond to a specific person you'd find in the data — it's a statistical computation.
- It complements, doesn't replace. The goal is to fill the gap left by privacy and consent, not to replace direct measurement.
- Model quality depends on the input. The better and more consistent the data from the measurable part, the more reliable the estimate.
The connection to server-side
Server-side tracking doesn't replace modeling but improves its input — by measuring the part that can be measured more reliably. The more accurate the direct data, the better the basis for estimating the rest.
Summary
A modeled conversion is a statistical estimate of conversions that can't be measured directly — a useful way not to lose the whole picture in a world of consent and privacy. Treat it as a complement to accurate measurement, not a substitute for it.