Modeling The Full Value Of Marketing Investments: Doing Precise Things With Imprecise Data?

  • by , Op-Ed Contributor, November 17, 2020

Day One of ARF’s “Attribution & Analytics Accelerator” event offered intriguing marketing modeling case studies from four major brands, Microsoft, Mazda, Trojan, and Cotton Inc., along with their analytics partners.  They revealed how they are embracing additional science and mathematical techniques with the art of understanding and optimizing their marketing investments from a truly holistic perspective to address an incredibly complex customer environment. 

The event is being jointly run by ARF and Sequent Partners.  Sequent Partners are acknowledged as the industry experts in this arena which includes marketing mix modeling (MMM)  Nielsen, the day’s sponsor, concluded the session with insights and a plea to manage marketing budgets long-term based on building and protecting brand equity even in such “unstable times.” Refreshing and so critical! 



Under the guidance of Session Leader Greg Pharo, global director, media analytics & advertising research, The Coca-Cola Company, the four very different brand modeling approaches presented underlined the difficulty of including and reliably sourcing the extensive data elements required of any valid modelling approach to understanding a brand’s marketing drivers and how to optimize them over time. 

Kathleen Magnuson — senior data and applied scientist, customer & market research, Microsoft reported on how the company has unified separate success measures (KPIs), derived from separate systems, managed by separate research teams. with a holistic approach.  Their need was to incorporate both short-term and long-term outcomes with a more granular model based on both direct and indirect effects.  In her presentation, “Realizing the Full Value of Advertising: A Granular MMM Success Story,” she suggested that they are now able to make “intense” data-based decisions about the right balance between brand-building advertising (long term effects) and promotion (short term effects). 

I’m not sure where their greatest weakness, in my opinion Customer Support, comes in the mix, but…

Mazda faces a classic multi-faceted marketing challenge: How to drive brand predisposition, motivate the consumer along the path to purchase at the showroom with a willingness to pay a premium price.  To address this complex puzzle, exacerbated by multiple layers from manufacturer to dealer associations to dealers, Mazda has synthesized analytic insights across carlines and markets to optimize resource allocations. 

Satya Menon, managing partner, ROI practice, Kantar, Brad Audet, CMO, Mazda and Eric Uchida, senior vice president, data, analytics & technology, Garage Team Mazda presented, “Generating and Disseminating Analytics Insights.”  This marketing approach is now helping to maximize the full value (ROI) of marketing for brand growth across upper and lower funnel strategies at the right time to drive business success for all parties.  Paralleling other brands, Mazda is clearly becoming a more data driven marketer via an omni-channel holistic customer knowledge center to ultimately drive dealer traffic.  A key enhancement that Mazda is now realizing is, “Integration of customer journey perspective in audience targeting.” 

Dan Bracken, vice presidentconsumer engagement, Church & Dwight, and Greg Dolan, CEO and Co-Founder, Keen Decision Systems, shared, “How Trojan Harnessed Its Full Marketing Horsepower.” Trojan’s marketing team is leveraging priors from attribution studies, household panel measures and store-level trade analyses to create what they believe is a comprehensive and complete view of the brand.  

Their emphasis was on building a more forward-looking predictive analytics model to establish a continuous optimization system which informs ongoing decisions.  Critical for any brand but especially a market leader that has 60% market share.  However, their marketing performance by media platform analysis may need a more detailed technical appraisal? 

Cotton Inc. has a unique marketing dilemma--No sales data, they are an association!  So Kim Kitchings, senior vice president consumer marketing, Cotton Incorporated, underlined its ROI proxies, including “check-the-label” and “intent to buy” in her presentation, “Measuring MORE of the Marketing Metrics that Matter.” 

Jim Friedman, CEO, Marketing Attribution Partners reviewed how the firm built a model that encompasses all the steps in the consumer journey to sales.  It demonstrated the sheer complexity and difficulty involved with meaningful, rigorous data capture, including direct and indirect and immediate versus longer term effects.  Yes, this is “hairy” stuff!  But it is hopefully no longer smoke and mirrors.  

Kim identified two sound conclusions from this on-going approach.  This kind of intricate detailed analysis improves segmentation of consumers.  It opens richer discussions with their agencies. 

With brand being one of our most valuable assets, finding the balance of protecting brand equity while driving immediate growth—all within a very uncertain global economy—is incredibly challenging. However, it is possible with the right mix of proven band + outcome metrics and normative benchmarks to fuel your full funnel marketing plans.

Tsvetan Tsvetkov, senior vice president marketing effectiveness, Nielsen offered the most thought-provoking presentations of the day – “Brand Stability in Unstable Times: Managing Your Marketing for the Long-Term.”  Built over many years of data capture and analysis, Nielsen’s resulting insights were used to remind the audience of the sheer importance of a marketing attribute which has often been lost in the social digital and sales worlds of “now” – brand equity.  He suggested that marketers can use past performance data and benchmarks to plan and make predictionsfor a very uncertain future.  Experience has demonstrated that achieving a greater share of projected growth can be achieved by understanding brand equity, the long-term effects of media and using normative outcomes and benchmarks.  Two cornerstones Nielsen has derived:  >10% of ROI variation is driven by socioeconomic factors; brand equity drives ~10% of sales.  In short, protect your brand by building brand equityby applying data to maneuver through uncertainty. 

In what may be a recurring theme over the next 3 days, “media” was often identified as a key ad investment/driver variable along with media ROI’s.  It is noted that in the US, with the exception of Out-of-Home (GeoPath markets) and Radio (PPM markets), “viewing,” “hearing” or audience ad “exposure” is not measured and consequently has no data source. 

And yet at the recent asi Conference in London, Karen Nelson-Field of Amplified Intelligence wisely stated, Without viewing (“seeing” or Eyes-On) there will be no chance an ad can work.  Mike Follet, Lumens, emphasized, It is unreasonable to ask media to guarantee an outcome as they would be taking on risk that they neither control nor are accountable for.” 

While these approaches have made great improvements and become more sophisticated according to Sequent Partners, perhaps they are still doing precise things with some rather imprecise (or misunderstood or misrepresented) data?  Stay tuned for my Commentary on Day 2. 


3 comments about "Modeling The Full Value Of Marketing Investments: Doing Precise Things With Imprecise Data?".
Check to receive email when comments are posted.
  1. Jim Spaeth from Sequent Partners, November 17, 2020 at 3:24 p.m.

    Tony - Thanks for the in-depth coverage of Yesterday's sessions.  I found them enlightening and encouraging.  

    You are right, of course, that there are no people in these anaytical systems and that's a challenge for marketing that, at it's best, calls on emotionalal appeals.  But even when we use Nielsen-sourced GRP data as inputs for marketing mix models, as we have for for over three decades, we still have no insight into what motivates consumer response.  We just see the media delivery of advertising, the sales response and the concommitant ROI.

    Today we are working at a much more granular level.  But it is not much different.  We have an impression (with or without an attentive consumer) and a sales response.  The inner detail of who was exposed, if anyone, how attentive they were, and their predisposition to respond, are all missing data that leaves us without explanation for the success or failure of the advertising.  But we still can measure the success or failure.

    We wish it were different, but you have to remember that this enterprise requires samples in the millions.  I can't imagine that the media world would pay for opt-in panels of registered individuals at that kind of scale. So, as often happens, we have to work with what we have.  And I think the marketers and modelers we showcased yesterday are doing brilliantly!

    Stay tuned ... we have two more exciting and informative days of the Attribution & Analytics Accelerator!

    Jim Spaeth

  2. Alice Sylvester from Sequent Partners, November 17, 2020 at 3:27 p.m.

    Thanks for the deep overview, Tony.

    I think one of the big take-aways from the day was the enormity of the effort required to wrangle all the media and marketing data available today into something analyzable and usable. The modelers are doing extremely impressive jobs injesting, cleaning, standardizing disparate data from many, many sources. This becomes even more important when the goal is to broaden financial performance evaluation beyond short-term effects driven by traditional media.  

    Another area that was covered was how to win organizational adoption of analytics that don't always support what people have done before and may contradict widely-held beleifs.  Multidisciplinary analytic teams, with strict data governnance practices and disciplined reporting structures are the keys to achieving success.

  3. Ed Papazian from Media Dynamics Inc, November 17, 2020 at 4:32 p.m.

    Tony, Jim and Alice, I agree to a considerable extent with all of you. As Jim states, there is no chance that advertisers would fund an attempt to refine the data represented by millions of actions so they reflected eyes-on-screen attentiveness to commercials---even if this was possible---which I doubt. However, there is also no doubt in my mind that using "raw audience" data---unadjusted for real ad exposure levels---will produce dubious results or, at best, mask the true effects of ad expopsure upon sales or purchase outcomes.

    So why not establish some normative eyes-on-screen values for various types of TV exposure---by show type, daypart, etc. from a source like TVision and model these into the data to take a look at what happens? For example, say one of the investigations concerns the effects of frequency levels. In addition to exploring what the "granular"  but "raw" data has to tell you, why not factor in various real attentiveness indicators, using probability assumptions and compare the findings? One dataset---the granular one---might suggest that the optimum frequency over a four -week period is 6-8 but the alternative finding, based on probable eyes-on-screen attentiveness, might shrink this to 3-4. That would certainly give everyone food for thought and might lead to the industry eventually insisting on using eyes-on-screen measurements in conjunction with "raw audience" data to get a more realistic picture of the imact of various ad exposure variables for ad campaigns.

Next story loading loading..