Over the last 10 years, digital media has been transformed in both quantity and quality. Facebook started advertising and acquired Instagram. Twitter, Snapchat, and LinkedIn followed suite and most recently TikTok. Not to mention Google, Bing, Yahoo, and their display networks. The increasing complexity of digital channels and formats, each serving one or more advertising purposes, means that a single user will probably be exposed to multiple ads before they eventually make a conversion or a transaction. That fact brought about a crucial question: which of these ads influenced the customer’s buying decision? In what platform and/or format should an advertiser invest the most? And, which ones should they leave out of their media mix? These are the kind of questions attribution models can answer.
Previously, marketers used to depend on measurement, down to the individual impression and consequently determining the return on ad spend (ROAS). But since this is no longer good enough, there was a need for “Attribution”. An attribution model is a method of distributing credit for conversions among the different ads & touchpoints. The default model used to be last-click attribution. But that model is neither fair nor realistic. Take for instance a school project. Those projects are usually done by a small group of students. This group may choose to assign the responsibilities in such a way that one person collects the data, another one tracks down the sources, a third is responsible for the write-up and a fourth can take on the presentation of the final submission. How would the professor then grade each of them? Supposing that he follows the last-click attribution model, the guy who did the presentation should get all the credit for the project and his other collaborators would not get any marks, which doesn’t make much sense since the final submission would not exist if not for their contributions.
Such concerns gave birth to multi-touch attribution models. A position-based model - for instance - would give more credit to first and last touchpoints than the ones in the middle, while a time-decay model would assign the credit to touchpoints based on their recency. Then again, these are one-size-fits-all kinds of models, that do not necessarily fulfill the requirements of all businesses, if any. That drove businesses to create custom models that would best fit their business needs. Later, and thanks to the strides made in the artificial intelligence field, many providers started offering data-driven attribution models that evaluate every single conversion path and compare it against similar ones to assess the contribution of every touchpoint. These elaborate models have changed the way advertisers and marketers look at their advertising. It gave more importance to channels & formats that were previously neglected and brought back some of the overappreciated ones to their true value.
But can we really trust these models? This question is a bit tricky to answer. Many of the attribution providers are publishers themselves (e.g. Facebook & Google Ads). This means that publishers might inflate their own contributions to a certain extent and downplay that of other platforms. Advertisers, therefore, turned to ad servers and other third-party platforms for a more fair representation of their media mix, meaning that they would need to share their advertising data with more platforms amid rising privacy concerns, not to mention the issue of the walled-gardens who don’t like sharing their proprietary data. So, rather than asking what the best attribution model provider is, an advertiser should rather ask what provider and/or model best satisfy my business needs.
In summary, attribution is crucial to understanding the true value of your ads whether digital or any other trackable media. Getting it right is a project worth investing time, effort, and money in.