Joe Stevens
The world of digital marketing is doing the one thing that it’s reliably done since it’s inception … change.
Most recently, changes are resulting from increasing consumer consciousness of how data is being used and subsequent data-privacy regulations. From GDPR to the anticipated reforms to the Australian Data Privacy Law.
As a result, digital marketers are getting modelled conversion outputs far more often than they used to, especially from sources like Google (who are integrating AI powered modelling) and Facebook (who have modelled conversions for some time) even with solutions such as enhanced conversions and CAPI enriching the data that digital marketers do have access to.
As our previous last click conversion attribution signals diminish many marketers are harking back to a technology first developed in the 1960’s-1970’s using Media Mix Modelling (MMM) for answers. The benefit of this has been has been to unearth actionable insights while staying privacy compliant, allowing for greater collaboration between above the line and below the line worlds by giving more parity to above the line when evaluating the ROI of specific campaigns, products or business units and shifting to proactive decision making by analyzing longer term patterns and trends.
Many MMM’s are even becoming more and more effective at forecasting, allowing those insights to be modelled for their potential effects and measured against them.
We’ve seen the benefits of using an MMM play out across a number of key clients at Atomic and I’m a massive advocate for their use.
Media Mix Models are allowing the opportunity to integrate more and more granular data into the platforms. This allows pinpointing and testing of the effectiveness of marketing tactics such as ad formats, buy types, placements and audience targeting. However, my advice would be to proceed with caution.
The beauty of having a more holistic overview of the performance of media is that it allows for more strategic shifts at regular intervals over a long term view and facilitates the kind of big adjustments that can be massively beneficial for business goals.
These include challenges such as: What channel mix is most effective?
What levels of investment are optimal for each channel?
Are we even getting ‘skin in the game’ in this channel with this level of investment? Which channels are we overinvested in?
How has our marketing efficiency increased over the past year?
Which channels and suppliers should activity increase in?
Big, meaty, questions. What, in my opinion, should be guarded against are items which would have traditionally been A/B tests in the domain of Performance marketing trying to be ‘ported over’ to MMM from domains such as in-platform tests and Google Analytics. Many of these tests are either too granular or too short term to return meaningful results in a media mix model (thank goodness) but MMM’s and the metrics such as yield and MROI are too often being leaned on to provide the answers.
Yes, these tests are important, but not all roads point to an MMM (just as not all 1 lane gravel tracks should feed onto a single highway).
If the data feeding into MMM’s becomes too granular and the analyses of the data within them become too short-term in nature we are misusing the platforms and risk not being able to answer the bigger questions and discussions will be far ‘smaller’ and less impactful than they could be.
Your marketing insights vehicle will become stuck behind tractors, heavy haulage vehicles and those dreaded slow caravans (to really labour the unnecessary analogy).
So, I implore any MMM users, define the specific scope of these hugely insightful tools and stick to it. Don’t make your media mix model Google Analytics 2.0, your key stakeholders will thank you for it!
Joe Stevens is client lead of performance at Atomic212.