After Day One of ARF’s “Attribution & Analytics
Accelerator” I suggested that based on the cases presented, Attribution modelling could be doing precise things with some rather imprecise (or misunderstood or misrepresented) data, notably
media data and especially all the discordant television data.
Day Two appeared to confirm that position as data challenges of every dimension and how they are being addressed,
integrated, modelled and used became clear.
Cadillac, Regions Bank, CIMM, ANA/AIMM and the day’s sponsor Analytic Partners shared their experiences and how they are dealing with
the complex, multi-dimensional data issues and its harmonization to holistically address cross platform attribution and its assessment.
Guiding brand success via strategic brand planning
for both the short and long term based on analytics is paramount for advertisers and their agencies.
But the question appears to be, is the data any good?
The Session
Leader, George Ivie, CEO & executive director of the Media Ratings Council was well chosen considering MRC’s critical industry role that is expanding into this arena. MRC will ultimately
ensure that the quality of all dimensions of “Attribution and Analytic” approaches is assessed based on on-going technical appraisals and evolving “Outcomes Measurement”
Standards. These standards are already being developed by an MRC Working Committee. (Full disclosure I am a member of that Group.) As Ivie stated, “What is achievable via
better practices?”
Eric Neville, senior manager, Media & Audience Strategy, Cadillac addressed, “Driving Campaign Effectiveness in a Cookieless World” with Angelina
Eng, vice president, Measurement & Attribution, IAB & IAB Tech Lab. It revealed just how much Cadillac (and no doubt GM) know about their consumers at the individual level. Their
goal is to own all the relevant data on their owners, prospects, and competitive manufacturer model target groups via extensively integrated databases. This provides the opportunity to profile
key segments and address them via optimum channels.
Cadillac consequently takes an audience-first approach to audience activation via TV and other media platforms. Based on the
media “exposure” definitional issue that evolved, their definition of media ‘impressions’ reported would be insightful.
Regions Bank have been using Marketing
Evolution for many years to leverage an innovative solution to understand consumer attributes, cross-channel measurement and consumer-level effectiveness measurement while suffering from incomplete
data sets. Incomplete “exposure” (not clearly defined) reach and KPIs data as well as data with varied granularity stymied their measurement approach. During their,
presentation (“Cross-Channel Audience Impact Based Planning with Incomplete Consumer Data”), Alyssa W. Blair, senior vice president, regions corporate marketing, Regions Bank and Michael
Cohen, Chief Technology and Data Officer, Marketing Evolution, explained how their “new solution” enables Region Bank to build, recommend and target audiences based on media
“impact” without individual consumer identity resolution. However, it must be noted that impact primarily encompasses brand and creative power at a minimum rather than being a media
effect. Alyssa underlined the value of their agency partners in assessing the optimizations and performance data.
Howard Shimmel, President, Janus Strategy & Insights led,
“Getting Attribution Right: An Exploration and Best Practices for Television Data Inputs in Attribution Modeling.” This project was sponsored by the Coalition for Innovative Media
Measurement, CIMM, under the leadership of Jane Clarke, CEO & Managing Director. The issue? Different television data streams preferred by leading modelers frequently lead to
dramatically different results and business decisions. So, what really drives the difference in attribution results?
Regrettably, the term “exposure” took on a variety
of interpretations during this presentation. It appeared that much of the TV data currently being used in attribution models is technically described as “viewable impressions.”
This data only reflects whether an ad has been delivered and rendered to specification to a screen. Viewable impressions are neither a media audience currency nor any measure of
“viewing” by an audience with eyes-on or contact.
Of note, the accuracy of match rates of qualified viewable impressions spots to the TV brand schedule differed significantly
in this evaluation across several providers for the same campaign! In addition, Nielsen ratings data which is also used by some attribution vendors is also shy of measuring “viewing”
although it was described as, “exposure data.” (It is opportunities-to-see, OTS, or gross impression data.) Attribution is all about understanding outcome, of course.
However, if there is no audience “viewing” or contact, there can be no outcomes!
“Remove Blindfold Before Targeting!” presented an outcry for greater transparency
and data quality from ANA’s Alliance for Inclusive and Multicultural Marketing, AIMM. Cultural digital identity data has been a black box with unknown accuracy and coverage –
essentially systemic inequities and significant under representation. While this can be a challenge for targeting, it threatens the validity of multicultural attribution and analytics.
Carlos Santiago, Co-Founder, ANA/AIMM and President, Santiago Solutions Group and Aaron Fetters, Head of Client Development, shared the overview. Truth{set}’s benchmarking evaluates the
accuracy of individual personal data at scale and produces population segment accuracy indices. These can range from 0.93 for sex to 0.22 for Hispanic as basic examples! Good to know that
MRC is involved in this sensitive area which will increase transparency.
Katie deGorter, associate vice president, Analytic Partners evaluated, “The Next Evolution of Analytics for
Decisioning,” in the most forthright presentation of the day. She reminded the audience of the ongoing data apocalypse and what she suggested as, “the failed promises of multi-touch
attribution which is going to get worse.” How do organizations evolve measurement strategies in this new world with increasingly limited consumer tracking and data access, bias, reliance
on siloed and partial data that exacerbate the analytic risks due to inaccuracy and therefor potentially flawed insights? Analytic Partners embrace a holistic, integrated multi-dimensional data
approach. This basis, they believe, enables their commercial mix models (decision systems) to deliver the greatest value to brands as they do not require granular data to derive granular
marketing insights.
George Ivie’s summary and conclusions were absolutely on-point and need to be fully supported by the industry via MRC membership. “The quality of
source data is fundamental and critical. It is often inconsistent and non-harmonized, etcetera which makes assessment extremely difficult …How Attribution is done requires a much higher
level of transparency.”
For MRC this would entail a full and detailed Technical Appendix as warranted by any model, survey, or approach. “On-going test and control in a
disciplined manner is required.”
Stay tuned for my Commentary on Day Three – Analytics-Driven Business
Results featuring cases from another four major brands.