What are some advantages of a statistical attribution model versus a rules-based attribution model?

Rules-based models assign credit to one or more touchpoints according to rules that we set. Think of these models as different lensthrough which you can analyze results. For example, giving credit to the first and last touchpoints (impressions and clicks) leading to a conversion. Rules-based models using people-based measurement provide deeper insights than cookies-only measurement, since multiple touchpoints are tracked and correlated with an actual person.

With rules-based models:

You define “the rule”: How should credit be allocated?

Results are based on your model choice.

You choose where along the path to conversion you want to assign the credit: last click, even credit, time decay, or positional.

Statistical models use algorithms to determine credit for each touchpoint. These models are driven on a result (increased revenue) rather than an assumption of where to assign credit. They use all available data to determine which ads increase revenue. Like rules-based models, statistical models are far more effective at analyzing past results and predicting future results when using a people-based measurement tool.

With statistical models:

Algorithms define the credit allocation.

Results are dynamic and learn from historical data.

They're also referred to as algorithmic MTA, data-driven MTA, and results-driven MTA.