#224 - MMM is a powerful tool, but not a silver bullet

Well, today I think we wanna talk in terms of strategy, a really important part is measurement and media allocation.

I think you measure your paid programs and your marketing programs to understand what to do more of and what to do less of.

And so one of the-

Mm

... very common measurement vehicles that we've talked about in previous episodes and is really popular in the industry right now are three of my favorite letters, MMM, or a marketing mix model or a media mix model.

Yeah, Rich, for the uninitiated, why don't you tell us a little about that?

It's interesting, MMM, it's gone from initially like a bug word to almost like a mythical status in the world of marketing, especially for those industries like healthcare, the, where data is not as plentiful, and it's a sort of a level of sophistication that everybody is striving to get to, this gold standard.

It's a predictive model that is trying to, and you can also probably describe it a little bit more eloquently than I can, Ben, but essentially understand the effect that each channel is having in the marketing mix, and then using the historical data and using that model to then predict what you will drive with shifts in your media mix.

Even if you keep your media mix the same, but just based on things like the seasonality indexes that it's ingesting, it's essentially not just a more accurate or a prescriptive model of measuring in through a different lens, but also it's got this predictive element, which is why I think it's so fabled, and people love it so much, is because they see it as sort of this magic forecasting tool that's gonna tell them everything that they should do and how much money they should take out of Google and stick into Meta [chuckles] and how much money they should take from channel A to channel B. It's something that everybody wants.

There are obviously some limitations.

The old adage is, "All models are wrong, it's just that some models are useful." And I know in this podcast we're gonna get into some of the challenges around using MMM, but I think we like to think about it's a fantastic measurement solution as part of the measurement toolbox if you are eligible for it, if you have the right amount of data, if you have the right amount of spend, if you don't have huge amounts of volatility in your business is gonna make the modeling difficult.

And especially in a world where you may have pixel-based tracking constraints with HIPAA, the inability to, for these ad platforms to sign BAAs that require you to remove certain tracking from the website.

It's a great solution that sits inside of a broader measurement approach.

What it is not is probably a let's abandon every other way that we've traditionally done measurement and solely rely on MMM, because what you'll quickly find is you don't have the level of granularity and detail that an organization is often looking for in order to make the right thing across the board if you just put all your eggs in that single basket.

Yeah, for sure, and I think most of our audience on this podcast are coming with a healthcare lens, so I'd love to talk a bit more around what are some common challenges that folks in that industry face when trying to deploy MMM.

You mentioned HIPAA, you mentioned pixel tracking.

What are some common breakdowns or limitations or strengths that the healthcare marketing ecosystem can take better advantage of with MMM?

Yeah.

It's getting your hands on the data to begin with.

The good thing and the bad thing about MMM is that you don't need any website event data, you don't need any pixel data, which is a plus column for a lot of compliance-constrained clients.

You can use encounter volume, new patient volume.

You could use patient revenue.

There is a whole lot of downstream activity that you can use to feed the model and build that model off.

Now what you need though is you need ideally 22, 24 months of clean data, and you need a certain amount of data in order to get an accurate model, even at the channel level.

There's different points of view on this, and depending on what kind of MMM solution you leverage, different levels of sensitivity.