Conversion modeling is how Google Ads estimates conversions it can't observe directly, for example when browsers block cookies, visitors decline consent or people switch devices. Google uses data that doesn't identify individual users, reports modeled conversions only when it is highly confident they happened, and includes them in the Conversions column alongside observed ones.
Not every conversion can be tied back to an ad interaction anymore. Google fills part of that gap with modeling: it studies similar traffic it can observe and estimates the conversions it missed in the traffic it can't.
Why are some conversions missing?
Google lists these reasons:
- Browsers. Safari and Firefox don't allow conversion measurement with third-party cookies, and Safari limits how long first-party cookies last.
- Consent. Regulations in some countries require consent for advertising cookies, so visitors who decline aren't measured the normal way.
- Cross-device journeys. When someone clicks an ad on one device and converts on another, the conversion may not be attributable.
- Apple's App Tracking Transparency. Google doesn't use information covered by the ATT policy, so conversions from that traffic are modeled.
How does Google model them?
Google uses data from observable slices of traffic that behave the same as, or very similarly to, the slice it can't observe. For example, it looks at conversion rate trends across browser types, together with coarse dimensions such as device type, time of day, location and operating system. The data it uses doesn't identify individual users.
To check accuracy, Google uses holdback validation: it holds back part of the observed conversions, models that slice, and compares the result with what really happened. It only reports modeled conversions when it is highly confident they occurred as a result of ad interactions.
In your account, modeled conversions:
- appear in the Conversions column together with observed ones, at the same level of detail, including conversion value and attribution path;
- can take up to 5 days to fully process, and may increase retroactively for several days;
- flow into the reports and bidding that use conversion data.
How does consent mode affect it?
If you use consent mode, Google models conversions for the visitors who decline consent. To qualify, your account needs 700 ad clicks over 7 days, per country and domain grouping. The advanced implementation, which sends cookieless pings, gets an advertiser-specific model. The basic implementation gets a general, less detailed model.
After at least 7 full days with consent mode, Google Ads may report the uplift from modeling, but only for slices that meet the minimum data threshold. The details are in basic vs advanced consent mode and whether consent mode is mandatory.
How can you make it more accurate?
Google's advice is to give it more observed data to learn from:
- the Google tag on every page of your site;
- consent mode, so Google can model unconsented traffic;
- enhanced conversions, which Google says help where browsers limit first-party cookies.
In my experience, modeled conversions are worth trusting as part of your totals, since bidding uses them anyway. Where I stay careful is right after a tracking or consent change: if conversions jump the week you switch on consent mode, compare them with your backend orders or CRM leads before you raise budgets. I cover the setup in my consent mode v2 guide and in the step-by-step enhanced conversions tutorial. For which version fits you, see enhanced conversions for web vs for leads.
Sources
- Google Ads Help: About modeled online conversions - https://support.google.com/google-ads/answer/10081327
- Google Ads Help: About consent mode modeling - https://support.google.com/google-ads/answer/10548233
- Google Ads Help: About consent mode - https://support.google.com/google-ads/answer/10000067
