At Recast we're on a mission to rid the world of wasted advertising spend. No one method or tool is going to get us there, so as well as our own content, we're sharing the best of what we've read on marketing measurement this week.

Meta Robyn or Google LightweightMMM - which is better?

I’ll avoid the stock answer “it depends” and give you some context to help you decide. Both models are sophisticated and likely better than what you’re getting with an outsourced agency, who are often just manipulating the numbers to show you what you want. Automated MMM beats hand-crafted artisanal MMMs, because you at least know any bias is a result of the data and not whoever built the model.

We use Bayesian methods at Recast and as such would be naturally drawn to LightweightMMM, as a better foundation for building the various functionalities we’ve found important, like time-varying coefficients and plausible priors. That said Robyn is a great model and if you weren’t planning on any customization at all, it’s probably the best out of the box.

We try to read everything out there that's relevant to marketing measurement, and this is the best of what we've read recently.

If you read something great this week, or want to share your own content, submit here.

This is the best of what you might have missed, anything previously posted that is still relevant today.

Just for fun...