How conversion matching attribution works

Overview

Conversion matching attribution takes conversion events* recorded in the source of truth and matches them to the conversions reported by the digital marketing and affiliate platforms. The source of truth for conversion events varies from business to business but is usually one of the following: a database, CRM, commerce platform or ERP. A conversion is recorded by digital marketing platforms and affiliates when a user interacts with the marketing activity and subsequently completes an action that is registered as a conversion event/goal in the platform.

*A conversion event is any action a prospect/customer takes which generates revenue or increases the probability of generating revenue examples include a website visit, subscription, signup and checkout.

Disaggregating reported conversions

Unfortunately, the conversions recorded in the digital marketing and affiliate platforms are typically reported in aggregate (e.g. advert x had 11 conversions on the 23/3/2024). However it is possible to request these aggregated values sliced by a number of dimensions such as the hour the conversion took place, the hour the impression took place, the country the user is located in and many more to retrieve multiple reporting tables at varying granularities. From these tables an algorithm can be used to construct a table that is one row per conversion for each digital marketing platform. These conversion tables include all the retrieved dimensions of each conversion.

Matching Conversions

It is then possible to match the conversion events in the source of truth to the conversions in the conversions tables based on the criteria they have in common such as country, conversion type and hour. This is done using a probability matrix. It is frequently possible to establish a one-to-one match but occasionally the following instances arise:

• Platforms reporting the same conversion: One conversion event in the source of truth matches multiple entries across the platform conversion tables - In this case the conversion event is attributed evenly across the adverts/affiliates with matching conversions.
  Example: If three different adverts all reported a conversion with criteria matching an order in the database then each advert would get 33.3% of the order and its associated revenue attributed.
• Unknown underlying conversion: One entry in a platform conversions table matches multiple conversion events in the source of truth - In this case an even proportion of each conversion event is attributed to the advert/affiliate with the matching conversion criteria.
  Example: If the criteria of a reported conversion matches three different orders in the database then 33.3% of each order and its associated revenue would be attributed to the advert.
• Platforms potentially reporting the same conversion and unknown underlying conversion: Multiple conversion events in the source of truth match multiple entries across the platform conversions tables - In this case the true overlap remains unknown so an average attribution is allocated.
  Example: If three different adverts all reported a conversion with criteria matching four different orders in the database then the true number of orders with an advert interaction is unknown. It is known that at least one of these orders did not have an advert interaction but it might be that they are all reporting conversions for the same underlying order, or that they are each reporting a conversion for three separate orders. In this case each advert would get between 25% and 75% of each order depending on how frequently these adverts have overlapping conversions.

Why conversion matching attribution overcomes the issues faced by ad platform reporting

• No double counting conversions - Conversion matching attribution uses conversion data from your own database/commerce platform meaning there is no issue with double-counting conversions. In the instance where multiple reported conversions correspond to the same underlying conversion event, the conversion event is evenly attributed.
• New vs existing customers - The source of truth records whether the conversion event was completed by a new or existing customer. Since reported conversions are always matched to conversion events in the source of truth, the issue of the digital marketing platform wrongly reporting that the conversion was completed by a new customer is eliminated.
• Highlights potential overlap - Conversion matching highlights digital marketing activities that are likely to have ‘overlap’. Overlap between two marketing activities is the extent to which users who interact with one activity also interact with the other. If the same underlying conversions are frequently attributed to both marketing activities this indicates overlap. It is then possible to see how overlap impacts metrics such as average order value and re-purchase rate.

Issues with conversion matching attribution

• All engagements do not have equal effect - Some engagements with marketing activity significantly increases a prospective customer’s propensity to convert (they are important) whilst others have no effect on propensity at all (unimportant). Conversion matching splits attribution evenly across all activities that report a conversion with matching criteria, meaning no distinction is made between important and unimportant engagements.
• Missing certain conversion events - Platforms only report on the conversion events configured against the marketing activity. In many platforms this is limited to one or two events that are frequently tied to the optimisation of the advert. This leads to many conversion events not being configured and therefore aren’t reported.
• Phantom conversionsConversions are recorded when a user takes a particular action such as visiting an order confirmation page or clicking a subscribe button. Sometimes the conversion action is triggered without the underlying conversion it is trying to represent taking place (e.g. visiting an old order confirmation page triggering an order conversion).
• Missed conversions Some conversions are missed because they are recorded using a tracking pixel which is a client-side code snippet meaning it can be blocked by users and browsers or they fall outsied of the conversion window.

Despite these issues, conversion matching offers an effective means of attribution. If implemented properly, marketeers can use the results to compare the performance of a specific activity over time or compare the performance of like-for-like activity. Which in turn enables rapid experimentation.