Did the Change Cause the Difference? Difference-in-Differences

How it works

The analysis splits your data into the classic 2x2 (treated vs control, before vs after) and estimates the change's effect as the treated group's before-to-after movement minus the control group's, via an OLS regression whose group-by-period interaction term is exactly that difference. It reports the estimate with a 95% confidence interval and p-value, in outcome units and as a percent of the counterfactual level. When a unit column is mapped, observations are first averaged to one pre and one post value per unit, a common way to limit the serial-correlation problem that repeated measurements create for the significance test. When the period column carries dates with several pre-change periods, it also tests whether the two groups' pre-change slopes differ, the parallel-trends check.

Use it when something changed for one group but not another (a pilot, a price change, a policy) and you have the outcome measured both before and after for both groups.

Not for a single group measured before and after (use the paired comparison or event impact tool. There is no control to difference against), and not a substitute for a randomized experiment when you can run one (use the A/B test tool).

What data do you need?

One row per unit per period, with the outcome, the group, and when it was measured. For example, weekly sales for pilot and comparison stores:

Minimum 12 rows · Best with 100-10,000 rows across 2+ pre and 2+ post periods

What's in the report?

Standard-library analysis: did the change cause the difference? Map an outcome, a treated/control group column, and a pre/post period column (labels or dates), and get the classic 2x2 difference-in-differences: how much the treated group moved, how much the control group moved anyway, and the difference between those two changes with a confidence interval and p-value. When your period column carries dates with several pre-change periods, the report also draws the treated-vs-control trends and runs a parallel-trends check: the assumption the whole causal read rests on.