Does old data influence current estimates?

Recast typically uses two to two and a half years of data when building your model. We’re often asked how old data is going to effect current estimates, since there’s usually a strong belief that the state of the marketing program two years ago was not reflective of what it was today.

Unlike traditional MMMs, the Recast model flexibly estimates ROIs that change over time. ROIs are allowed to drift up and down over time, but there are constraints on how fast they can change. The ROIs follow two built-in rules: yesterday should be similar to today (no major jumps) and seasonal patterns repeat themselves annually (ROI that trended up last spring will likely trend up again this spring). The data from two years ago largely helps us to estimate the seasonal component, but doesn’t strongly influence the current ROI estimate, since that has had such a long time to drift from where we estimated it two years ago.

You can see how the ROI estimate has changed over time on the Channel Level Deep Dive tab on your dashboard.