Media Mix Modeling has been receiving a lot of attention during the past couple of years, as a result of privacy related changes in the technology world. 
Privacy regulations such as General Data Protection Regulation (GDPR), Intelligent Tracking Prevention (ITP), App Tracking Transparency (ATT), and coming changes such as the Chrome Privacy Sandbox, Apple’s Privacy Manifest, and others, means that  Advertisers lost the ability to track users, and as a result, lost the ability to attribute individual conversions to the publisher that touched the user last.

Privacy related changes sent Advertisers into a measurement frenzy, searching for alternative methods to measure Advertising effectiveness, without relying on user level data.

The Resurgence of MMM

A method brought back to the spotlight was Media Mix Modeling, or in short: MMM

MMM, invented in the 1950s’ , was an econometrics regression method, relying on correlation of advertising spend and sales, with some adjustment, to measure the contribution of various marketing tools over sales figures.  

Advertising in the 1950s’, 60s’, 70s’ was somewhat different from Advertising today.
There were a limited number of channels to Advertise across, campaigns ran for longer, changing creatives, pacing, budgets, bids, and so on were, obviously, not possible.
 

MMM is a superb method of marketing measurement for when changes in the marketing activities are limited, and infrequent, which is why, MMM was the ultimate method of measurement for Advertisers during the years 1950 – 2000
 

MMM always had certain limitations

While ideal for retailers during the 1950s onwards, MMM carried certain limitations which a statistician or economist would need to address:
 

• Reliance on Incrementality experiments: 
  MMM could not differentiate the level of contribution each individual channel brought without conducting an incrementality experiment. Stopping a channel or a medium for a period of time, relaunching it, was a common exercise that Advertisers would need to go through for the sake of measuring the incremental contribution of each channel. This process was referred to as MMM Calibration.
   
• Parallel channels received the same results
  If two channels were active in parallel to one another, an MMM would think that both channels generate exactly the same results, even if one had a spend of $100 per day, while the other had $10,000 per day.
  To solve such issues, advertisers would often adjust the results by introducing a factor, thus, skewing the results to their liking. Similar adjustments had to be made when setting MMM Priors.
   
• No Organics
  MMM attributes all sales to Advertising results. In an MMM report, there are no organics.
  Introducing external contributing factors such as competition, holidays, special promotions, price changes, and so on is an extremely difficult task to include within the media mix modeling process.
  
  These limitation and others are why MMM was almost forgotten about during the 2000s’
   

Here comes Attribution

The era of digital Advertising introduced advertisers with the ability to track users. An Advertiser was able to drop a cookie, or collect a user id, and track the users’ journey until a user became a customer.

Attribution attracted Advertisers as it provided a sense of control. It worked in real-time, and allowed Advertisers to optimize campaigns, bids, budgets, and creatives regularly.

The biggest flaws in attribution had was that it would give 100% credit to the last publisher, often leading Advertisers to believe that certain channels had no value, simply because those received no credit. 



Privacy related changes made user level attribution lose its credibility, which is why Advertisers looked into alternative methods of measurement, giving a resurgence to methodologies like MMM.

 

Always on incrementality is an alternative to MMM

Advertisers today make more changes than ever. Two decades of digital advertising instilled new processes of daily optimization, creative iteration testing, bids adjustments, and so on. 
Each of these changes can be alleged as a sort of Micro Experiment.

While legacy incrementality measurement required geo lift experiments, or user a/b testing. And while user a/b testing is no longer possible given the changes surrounding user level privacy – GeoLift experiments required time, resources, and provided inconclusive results.

Tracking small marketing changes, allows Advertisers to completely flip the question of marketing measurement. Rather than trying to understand “Why did Jane download my app?”, Advertisers can now answer: “The increase in app installs was a result of the budget increase on our Google campaign”. All without needing to collect user level data.


 

More and more advertisers realize that while Attribution is a nice to have – marketing measurement should focus on measuring contribution. The value generated by each and every campaign, and channel, despite there being dozens of changes and optimizations happening simultaneously.

INCRMNTAL’s platform logs marketing changes automatically, reporting the incremental and marginal performance of each channel, and every campaign, without requiring user level data or conducting planned experiments.

Orchestrating Measurement is the name of the game

Attribution, incrementality, and MMM are not competing methodologies. They are just different methodologies. Smart companies will make use of the 3 methodologies, answering different questions, zooming in and out of their marketing activities.

While attribution provides the most micro-view of results – it will often not answer the question of contribution.
MMM is a great methodology for predicting the future, but it is an assumption based method, requiring substantial adjustment, manipulation, and bias.

You can read more about how to make the three methodologies of measurement play nicely together in the white paper we’ve recently published: Marketing Measurement Orchestration.