They want to know how long after an intervention there is an effect

interrupted time series (with a delayed effect and) some confounding.
Can covariates control the confounding?