I mean, you don't need too many stat techniques for basic tracking:

1) Set up Facebook pixel on your checkout thank you page + collect referral into into a database.

2) Spend $1,000

3) See how many purchases Facebook reports came from Facebook, check out how many sales you get recently have the Facebook referrer.

4) Repeat if sales revenue > $1,000.

This is obviously the simplest use case possible, but there you are. It's pretty reliable.

This is missing some testing. In multichannel environments you have the potential for multiple touches, as well as baseline if your company is large enough. For example, a traditional product like cars: you have a baseline of folks who come in, and then you have (ideally addtional/higher rate of) folks that come through the digital channel, such as engaging an online salesperson. Do the digital channels generate a higher rate of purchase than the baseline channel? If so, is the lift statistically significant? These are what I mean. Revenue moving through a digital pipeline could be because people happened to fall into that bucket but were already going to purchase anyway. In other words, some of the $1000 went to non-incremental marketing. The goal is to make sure the incremental marketing exceeds the spend.

I guess for (4) you mean, "repeat if [attributable] _profit_ increase [vs control] > $1,000"?

Not much point in driving $1,000 in _revenue_ with $1,000 in marketing spend unless your margins are 100%.

Thanks sib. That's exactly the point I was trying to drive home.

yes