Modeling Deep Dive
Below is a high-level overview of how Rockerbox builds a multi-touch attribution model for you.
- Rockerbox categorizes all marketing touchpoints into a tier 1-5 taxonomy
- Rockerbox aggregates these touchpoints for all website visitors, building a “timeline” of sorts for all users
- Rockerbox groups users by whether or not they converted
- Rockerbox runs a logistic regression to find which marketing channels (by taxonomy) are most highly correlated with converters
- This produces weights for each marketing channel taxonomy, which are then used to determine how much credit each channel receives
- In the MTA reports, Rockerbox applies these weights to all the marketing touchpoint rows & then they are normalized (so weights add up to 100% for each converter)
If you’d like to review a more in-depth technical explanation of this process, check out our guide here.
When you make significant changes to your marketing mix, Rockerbox can refresh your Multi-touch attribution model so that the model weights are reflective of optimized performance or new channels.
Reasons for a model refresh might be:
- New channel launches
- Significant changes in tactics within an existing channel