ABOUT ROCKETSHIP HQ: Website | LinkedIn  | Twitter | YouTube

KEY HIGHLIGHTS

⛳ How data volume is central to the effectiveness of media mix modeling

⏰ Media mix modeling is not always a good use of your time

🖖 The simple solutions will beat overkill every time

FULL TRANSCRIPT BELOW

An advertiser recently asked me: “I’ve been hearing about media mix modeling to address the very broken tracking on SKAdNetwork. Should we start evaluating media mix modeling?”

Now: they’re spending low six figures a month – on 4 channels. This ask brought up one of the frustrations I have about some of the advice on marketing and growth – the assumption that there are one-size-fits-all solutions.

Yes: last-touch tracking is problematic, and most digital measurement on multi-channel media mixes does not account for incrementality (ie: two of your sources might be targeting the same users) – especially with the broken tracking on SKAdNetwork. 

And yes: media mix modeling can be powerful. But it’s not right for the vast majority of advertisers out there. To state categorically that media mix modeling is the future (or for that matter: to state that any specific tool or methodology is the future) – without specifying who it’s right for, is incomplete – and possibly misleading.

So: why is media mix modeling not right for the vast majority of advertisers?

This is primarily because in order to get a media mix model that is useful, you need to have significant quantities of data about very, very different variables that drive your marketing outcomes. If you are P&G or Nike (or a digital advertiser that is spending in the tens of millions monthly), you likely are operating on multiple channels – social, search, other digital, programmatic, influencers, and offline. Each of these can influence your final revenue differently – and the more spend, click and purchase data you have the more powerful your media mix model will be.

Now: for the vast majority of mobile app advertisers (or for that matter, for the vast majority of offline or non-mobile advertisers), this quantity of data just isn’t available. 

And to be honest, if all you’re running is Facebook, Google, Apple Search and a few other programmatic channels, you really don’t need this level of analysis – especially if your first purchase, free trial or event that can be a somewhat strong predictor of LTV happens within the first 24-48 hours of an install (unlike many consumer goods products where purchases can be much more distant in time from exposure to media).

A related factor is that you need significant analytical sophistication to pull off media mix modeling that is actually helpful or useful. You need systems and team members to help you understand the interrelationships between different marketing channels and external factors.

According to an article in the Harvard Business Review:

Working with the vast quantities of data collected and analyzed through the attribution process, you can assign an “elasticity” to every business driver you’ve measured, from TV advertising to search ads to fuel prices and local temperatures. (Elasticity is the ratio of the percentage change in one variable to the percentage change in another.) Knowing the elasticities of your business drivers helps you predict how specific changes you make will influence particular outcomes. If your TV ads’ elasticity in relation to sales is .03, for example, doubling your TV ad budget will yield a 3% lift in sales, when all other variables remain constant. In short, analytics 2.0 modeling reveals how all driver elasticities interact to affect sales.

How many mobile app advertisers are able to measure the elasticity of TV ads in relation to fuel prices? Very very few, I might say. In another section of the same article, the authors say:

Crunching the vast database of driver elasticities, optimization software generates a set of most-likely scenarios along with marketing recommendations to achieve them. The software also can test specific what-if scenarios: For instance, how will sales of our midsize pickup truck in Denver be affected if gas prices climb 5% and we launch a combined TV and online campaign promoting a $300 rebate?

While there certainly are digital products, including apps with tens of millions in monthly spend that could benefit hugely from this sort of modeling, for the vast majority of mobile app advertisers that are operating off an MMP and/or an in-house database, this sort of rigor and analysis is both unrealistic and useless, not to mention a complete overkill.

What’s also challenging is that it’s hard to do basic MVP versions of media mix models – you could build spreadsheet models for incrementality, but it’s much much harder to build media mix models without significant sophistication.

While media mix models can be hugely powerful for advertisers who have the quantities of data, heterogeneity of media sources, quantity of ad spend, and sophistication of analytics resources to make it work for them, this isn’t right for the vast majority of advertisers out there.

Oftentimes, the simpler solutions are the best.

A REQUEST BEFORE YOU GO

I have a very important favor to ask, which as those of you who know me know I don’t do often. If you get any pleasure or inspiration from this episode, could you PLEASE leave a review on your favorite podcasting platform – be it iTunes, Overcast, Spotify or wherever you get your podcast fix. This podcast is very much a labor of love – and each episode takes many many hours to put together. When you write a review, it will not only be a great deal of encouragement to us, but it will also support getting the word out about the Mobile User Acquisition Show.

Constructive criticism and suggestions for improvement are welcome, whether on podcasting platforms – or by email to shamanth at rocketshiphq.com. We read all reviews & I want to make this podcast better.

Thank you – and I look forward to seeing you with the next episode!