Talk to them about “uplift modelling”. Most of the time, the campaigns will be estimated through some AB-testing. It could be a before and after test, or somet

Talk to them about “uplift modelling”. Most of the time, the campaigns will be estimated through some AB-testing. It could be a before and after test, or something even less significant like attribution. All of those have one thing in common: you assume that either you run a campaign, or you don’t. Given week enforcement, being louder wins. If one the other hand you decide to only send that email to people with a high chance of being interested and changing their behaviour, you limit your messages to a much smaller, more relevant subset. There typically is some pushback: “Why not also try people with a less than 0.1% of responding?” In that case, you want to introduce a cost: the fraction of a penny is costs to send an email, chances of being blacklisted, legal process, etc. Even with a conservative estimation of that risk, most of your targets become not worth it.