Modelled conversions are conversions that the system did not measure directly, but estimated based on the behaviour of comparable users. They emerged as a response to the limits of measurement – declined cookie consent, cross-device transitions, tracking blocking – where a part of the result would otherwise be lost entirely. In practice this means the figure in the report is a mix of measured and estimated values, which has two consequences. The first is that totals from the advertising system and from analytics will never match exactly, and looking for an exact match is a waste of time. The second is that the accuracy of the estimate depends on data volume – it is lower for small campaigns. Treat modelling, therefore, as an aid for trend and direction, not as precise record-keeping, and base important decisions on an experiment.
See also: Data quality in analytics, Incrementality, Offline conversion import.
