Conversion modeling

Conversion modeling is how Google Ads estimates conversions that happened but could not be measured directly, for example when a user declines cookies or a browser blocks them. These estimates, called modeled conversions, are added to observed conversions in the Conversions column.

What is conversion modeling?

Google defines modeled conversions as conversions estimated with data that does not identify individual users, for the part Google cannot observe (Google Ads Help, About modeled online conversions, accessed October 2026). In most cases Google still sees the ad click and the conversion; what is missing is the link between them. The model predicts whether the ad interaction led to the conversion. It does not predict whether a conversion happened.

How conversion modeling works

The model learns from the slice it can observe and applies that to the slice it cannot. With Consent Mode, Google uses the journeys of users who accepted cookies to estimate the conversion paths of users who declined (Google Ads Help, About consent mode modeling, accessed October 2026). Advanced Consent Mode sends cookieless pings, which gives Google an advertiser-specific model; basic Consent Mode only gets a general model (Google Ads Help, About consent mode).

Google also models other gaps: Safari and Firefox blocking third-party cookies, Safari limiting how long first-party cookies last, iOS traffic under Apple’s ATT policy, and cross-device conversions. Modeled numbers feed straight into Smart Bidding and can take up to 5 days to stabilize.

Conversion modeling data thresholds

For conversion modeling through Consent Mode, Google Ads requires a correct Consent Mode or IAB TCF v2.0 setup and 700 ad clicks over a 7-day period, per country and domain grouping (Google Ads Help, About consent mode modeling). For example (hypothetical), a site with 60 clicks a day in Vietnam gets 420 clicks in 7 days, which is not enough. Once the threshold is met, the model still needs a training period.

Google Analytics 4 has a similar feature, behavioral modeling, which needs advanced Consent Mode on every page, at least 1,000 events a day with analytics_storage=’denied’ for 7 days, and 1,000 daily users with analytics_storage=’granted’ for at least 7 of the previous 28 days (Google Analytics Help, Behavioral modeling for consent mode, accessed October 2026).

Example: reading modeled conversions in the Conversions column (hypothetical)

A shop spends VND 54 million a month. After adding a cookie banner, it observes only 180 conversions, a CPA of VND 300,000. It sets up advanced Consent Mode, meets the threshold, and Google estimates 45 more conversions. The Conversions column shows 225 and reported CPA drops to VND 240,000. Those 45 conversions are estimates, so check them against real orders in your CRM before raising the budget.

Limits of conversion modeling

• Google aims to avoid over-reporting, so some conversions that really happened are still not counted.
• Accounts with few clicks or very few weekly conversions may not get modeling; basic Consent Mode only gets a general model.
• It does not measure incremental impact. To know whether you would lose orders by turning ads off, you need incrementality testing.

Pair it with enhanced conversions: the tag sends customer emails and phone numbers hashed with SHA256 to Google, which adds observed data (Google Ads Help, About enhanced conversions, accessed October 2026). There is no official Vietnamese guidance treating hashed data as de-identified, so only send data from customers who have consented (see our guide to Vietnam’s personal data protection law).

Conversion modeling vs attribution modeling

Conversion modeling estimates how many ad-driven conversions went unmeasured. Attribution modeling splits the credit for a conversion across touchpoints. See also multi-touch attribution models.

Frequently asked questions

Is there conversion modeling without Consent Mode?

Yes, for gaps caused by browsers, iOS and cross-device journeys. Conversions from users who decline cookies are modeled only when Consent Mode or TCF is in place.

Which metric can you use to sanity-check modeled conversions?

Use MER (total revenue divided by total marketing spend), because it uses real revenue instead of the Conversions column. See MER, CAC and LTV.

If you want a second look at your Consent Mode, enhanced conversions and Google Ads conversion numbers, contact Adtek.