Understanding the difference between Digital Attribution and Media Mix Modeling

Ben Hinson
4 min readMar 26, 2016


Across the media landscape, one challenge that faces agencies, vendors and clients alike is a way to holistically assess marketing performance and allocate media spend across channels in an efficient manner that will ultimately maximize ROMI (return on marketing investment). Channel-centric media agencies place a strong emphasis on optimizing tactics within channels: paid search, display, paid social, email and broadcast media (TV and radio). While it is smart to maximize the knowledge of specialists, it is equally, if not more important, to assess the best marketing mix on both a macro and micro basis — in essence to see the big picture.

To date there are two standard practices that address both sides of holistic marketing mix assessment: Digital Attribution Modeling and Media Mix Modeling. These approaches complement each other with assessment and optimization.

Media Mix Modeling

Media Mix Modeling (MMM) can be thought of as the big brother to Attribution Modeling. Traditionally, MMM is a top-down approach used to assess how to best allocate spend between channels on a macro level.

Why shouldn’t marketers determine how to distribute their media budget by running simple ROI formulas, or using cost per action measures (CPA) or just relying on their own opinions? The answer is that the marketing efforts of each channel are not homogeneous. Clearly there are several different ad formats, delivery mediums and creative types used across multiple channels. Offline factors can also influence performance metrics. Media Mix Modeling creates response curves for each type of marketing spend and then uses that information to determine the most efficient marketing mix. MMM’s key strength lies with its ability to analyze offline variables that affect media performance: market elasticity, marginal profit, point of diminishing returns, seasonality and how other media impacts consumer behavior.

When done properly, MMM takes historic data into account, and is usually performed once a year (although some companies run it on a quarterly basis). It can help marketers determine ROI from each channel, and predict expected returns from future media spend. In essence, MMM assesses the cause and effect of changes to the marketing mix. Traditionally, MMM has been the best fit for brands that focus more on offline channels, but those lines are blurring in today’s digital age as companies across the board integrate their offline and online media channels.

With that said, MMM does have its limitations, namely:

· The variables that inform the model can change after the modeling data is collected and analyzed, making the model obsolete for the long term (e.g. media budgets may change after the model source data is compiled).

· MMM does not measure long-term brand equity, hence it is limited to short to medium-term ROMI forecasting.

· MMM may not be the best tool to speculate investment levels for new products.

Attribution Modeling

If MMM focuses on a top-down approach through channel allocations, then Attribution Modeling (specifically digital attribution) focuses on a bottoms-up approach that involves more granular planning and optimization on a more frequent basis. A true attribution model goes beyond a basic path report that is available on many generic platforms (e.g. Google Analytics, DoubleClick, Omniture Site Catalyst, Webtrends, etc.). It will offer more robust information such as:

  • Identify the most influential conversion points in a consumer’s digital journey.
  • Utilize an advanced algorithmic approach to evaluate performance and automatically update results.
  • Understand with precision how media assets performed along the conversion path.
  • Enable intelligent optimization recommendations, such as how should marketers shift budget by tactic, ad group (SEM), publisher (display), placement (display), keyword (SEM), etc.
  • Allow for scenario planning and forecasting (e.g. what the expected KPI return will be if budget is shifted between channels).
  • Directionally assess the impact of digital efforts.
  • Allow marketers to integrate learnings and recommendations with DSPs (Demand Side Platforms) and DMPs (Data Management Platforms).

Having a properly vetted digital attribution system in place enables marketers to understand how their channels, tactics and spend are performing, and allows for planning optimizations to be made on a real-time basis.

Attribution Modeling also has its potential limitations based on the solution/provider chosen, including:

  • Offline partnerships may offer limited match rates, where offline data is matched to an internal data pool, depending on the particular vertical.
  • Certain pathing approaches may add bias to results. Last touch and first touch models are considered inaccurate by many experts today as they are subjective and do not take into account the multiple interactions consumers have with a brand.
  • In many cases, results may not account for missing critical offline influences, such as in-person referrals.
  • Weighting assumptions (if the methodology is subjective and/or not holistic) may bias the results.


Despite limitations, Media Mix and Attribution Modeling can add great value to an organization’s holistic marketing strategy. Both approaches allow for efficient macro and micro optimizations. They can help marketers make strategic decisions about budget allocations by providing deeper insights into customer behavior.

In the next post in this series we will take a deeper dive into digital attribution, and review how to choose a potential attribution partner.



Ben Hinson

I enjoy creating content, solving problems, sharing knowledge, learning about our world and celebrating others. Learn more at www.benhinson.com