Modeled Conversion
Direct answer. A modeled conversion is an estimated association between an ad interaction and a conversion where the platform does not have a clear observable link. Google Ads documents using observed conversions and machine learning to predict unobserved conversions, then integrating observed and modeled conversions in reporting. Modeled reporting is not an identity-level ledger, does not prove causation, and does not reveal the modeled share or accuracy of any specific account without current account evidence.
Modeled conversion, defined
Google Ads describes conversion modeling as using observed conversions to predict unobserved conversions without identifying an individual [google_ads_conversion_modeling]. The platform separates interactions with an observable conversion link from those without one, uses patterns from the observed group, and integrates modeled conversions into reporting where its documented confidence conditions are met. That is platform behavior, not a first-party record of a named customer or payment.
The word modeled is the load-bearing part of the definition. It describes an estimate produced under a platform method. It does not mean false, and it does not mean observed. The existing revenue tracking and attribution systems owner applies the broader boundary: attribution output is a model used to assign credit, while internal verification is required before a financial statement is made [fp_tracking_attribution_owner].
Observed, imported, and modeled conversions
| Evidence class | How it enters the measurement layer | What it supports on its own |
|---|---|---|
| Observed conversion | The platform has an observable link between an ad interaction and the conversion. | That the platform recorded the linked event under its configured definition. |
| Imported conversion | A downstream event is supplied later through a supported import path. | That a submitted downstream event was handled under the platform process; matching and attribution still have limits. |
| Modeled conversion | The platform estimates a link where a clear observable link is unavailable. | A platform-modeled reporting value, not an identity-level record or causal conclusion. |
| First-party confirmed event | The business system of record stores the CRM stage, booking, invoice, payment, or other defined state. | That the internal record exists under its definition; not that an ad or page caused it. |
These classes can appear near each other in an operating report, but they should not be described as if they have the same provenance. An offline conversion import originates in another system and returns through a documented platform path. A modeled conversion is estimated by the platform where the observable link is missing [google_ads_offline_imports] [google_ads_conversion_modeling]. A first-party confirmed event remains the stronger source for whether the internal business state occurred.
Advanced consent mode is one documented modeling context
Google documents advanced consent mode as loading tags with default consent states and, while consent is denied, sending measurements without cookies; those signals can support an advertiser-specific model. By contrast, in basic consent mode, Google tags are blocked and no data is transferred when the user does not consent; Google Ads conversion modeling is then based on a general model [google_consent_mode]. The documentation supports two distinct implementation contexts. It does not establish the consent configuration, modeled share, data handling, or legal posture of any specific site or account.
Consent state is therefore not something a public page can infer from a conversion total. Verifying the implementation requires the current consent configuration, tag behavior, account settings, and applicable records. Deciding what a business may collect or send belongs to its accountable privacy and legal owners; this page describes measurement architecture and offers no legal or compliance conclusion.
How to read a report that includes modeling
- Confirm which platform, report, date range, conversion actions, and attribution settings produced the number.
- Record the platform's current description of where modeling is used and what the report exposes.
- Separate the platform total from identity-level CRM, booking, invoice, or payment records.
- Reconcile a bounded cohort where shared identifiers and permission boundaries allow it.
- Describe any residual as unknown rather than assigning it to loss, fraud, campaign failure, or data error without evidence.
- Keep the model and first-party definitions beside the result so later readers know what changed if the totals move.
What modeled conversions do not prove
- They do not identify a person, customer, invoice, or payment represented by an estimate.
- They do not prove that advertising caused a conversion or that a campaign produced revenue.
- They do not reveal the modeled share or model accuracy of a specific account without current provider evidence.
- They do not establish leads, conversions, revenue, ROI, ROAS, traffic, rankings, or AI visibility for a business.
- They do not replace internal verification, first-party reconciliation, consent review, or operator judgement.
Where to continue
Use the Tracking, Attribution & Measurement Systems domain for portfolio context, the canonical tracking and attribution guide for the evidence ladder, tracking gap for an unreconciled boundary, and offline conversion import for a documented downstream return path. A Revenue Scan can inspect public signals only; it cannot read modeling settings or account-level reporting [fp_revenue_scan_boundary].
Source and evidence notes
-
google_ads_conversion_modelingGoogle Ads Help — How conversion modeling works. Limitation: Documents Google Ads modeling behavior only; no account-specific modeled share, model evaluation, accuracy, or business-outcome claim. -
google_ads_offline_importsGoogle Ads Help — About offline conversion imports. Limitation: Documents a separate platform import capability; no matching, attribution, bidding, or revenue result. -
google_consent_modeGoogle for Developers — Consent mode, Tag Platform. Limitation: Documents the basic-versus-advanced consent-mode boundary, tag behavior and modeling context only; no site-specific configuration, legal conclusion, or modeled-share claim. -
fp_tracking_attribution_ownerOmniLabs Systems — What are revenue tracking and attribution systems? Limitation: First-party systems architecture only; not an external standard, implementation proof, or client result. -
fp_revenue_scan_boundaryOmniLabs Systems Revenue Scan public page. Limitation: Supports the public-signal versus internal-verification boundary only; no account state, modeled share, or financial impact is established.