Headway Senior Data Scientist Interview Experience — Technical Screen on a Diff-in-Diff Notebook

They gave me two notebooks and then we went through the results in them. One was a linear regression and the other was a diff-and-diff. They picked two cities, and in Massachusetts they gave an incentive program to increase availability, $200 per week. The data covers 36 weeks: the first 24 weeks are pre-launch, and weeks 24 to 36 are when the incentive program is running there. So I said that at the start we would probably assume parallel trends, and we can check whether the availability before launch was consistent. Next, I said we would want to rule out anything else going on during that period, like a marketing campaign or a product feature change, that might change other factors. I also said we need to take the pre/post change in availability in Massachusetts and subtract the pre/post change in Pennsylvania, and that is the change in availability. I also told them that for bookings we should look at the change with the same method. After that, we would convert the bookings into incremental revenue, and then do a tradeoff analysis against the real cost to see whether it ends up revenue positive. But then in step four of the calculation, I asked whether we are only looking at the change? It seems there isn't really any statistical inference we can do here, right?