Did it actually work?
This isn’t in the product yet.
30-second version
You changed one thing. We compare what happened with what probably would have happened if you had not. The gap is the effect, and it comes with a range. Causal cards are on the roadmap.
Do it
- Name the change and the date. A launch, a price, a channel you paused.
- Keep a what-if. Another product you did not touch, or the trend from before the change.
What you'll see
A gel line for what happened, a dashed line for what would have happened, a sun band for the gap, and a readout with a range. Marked SAMPLE.
Under the hood
The house designs are interrupted time series, difference-in-differences, and controlled regression. Each has its own page. None of them ship in the product today. A causal claim will always show the estimator, the interval, and the assumptions that would change the answer.
When not to trust it
- Another big change in the same week breaks the comparison.
- Fewer than about eight weeks of history is usually too little.