Thread by @mduboe

1/ CAC is one of those metrics that can be easily manipulated — intentionally or not.

But it’s not quite as simple as zooming into paid vs blended CAC.

A few thoughts…

2/ Much has been written about LTV:CAC ratios & dangers of (a) being overly-generous w/ the “L”, & (b) using blended CAC in this context

Expressing acquisition economics in terms of payback (# [months] for contributing margin to pay back upfront acq cost) is more accurate, BUT…

3/ Emphasis on paid CAC implicitly encourages use of last-touch attribution, which has its own shortcomings

Highest-intent channels/audiences most likely to claim last-click credit, but also least likely to be incremental

4/ At the same time, it is imperative to understand how downstream quality of acquisitions from a given channel/audience/keyword evolve over time. ie all payback math should be granular to channel (at a minimum)

So some form of upfront attribution is important

5/ How to solve? Incrementality (lift) testing.

Conceptually simple but can be difficult/expensive in practice due to the need for holdouts (and % holdout necessarily higher for lower $ spend — thus, practice is most common with at-scale teams)

6/ that said, fb & google have reliable out of box tools for conversion lift studies (standard caveat about bias, but methodology checks out) — and methods like ghost ads / PSAs are increasingly common for display-based channels

7/ when companies express CAC in blended terms due to “paid halo”, lift testing can quantify that halo

if more practical to use simple attribution method for weekly reporting, fine, but calibrate to lift test results. It’s more accurate, & usually better than fancy MTA solns

Topic for another day: importance of understanding marginal CAC (and cohortizing acq economics!)