We all see in the trade press that this is the age of data-driven marketing. It is reported that as many as 90% of advertisers are now using programmatic advertising, that is, ad placement done via data and algorithms that are precision targeted an individual at a time.
Bob Lord, President of AOL Platforms, reported in Ad Age on the CES…
It’s all about the data.
“…a carefully considered data strategy must be a CMO’s top priority…With data coming at brands from every conceivable direction and array of consumer devices…Marketers need to be able to analyze all the data generated by campaigns, monitored by sensors and captured by devices holistically — across all channels — so they can make the best decisions possible about the messages they create…”
So if marketing is becoming increasingly data driven, I ask, what is the corresponding data revolution for marketing RESEARCH?
It is time for marketing research to start leveraging naturally occurring data and adopt a predictive rather than report card mindset to improve the enterprise’s performance. Consider how a data science mindset might transform the following areas.
Marketing productivity. As a user requests a webpage, the marketer’s goal is to serve the most relevant ad that has the greatest chance of response and to win that impression at the right price (or pass on the opportunity). And to make this work at scale so billions of impressions can pass through these rules.
So what does research reveal that helps us achieve this goal? Currently, very little.
We rely on regression models with aggregated variables where the user is not in the equation; the focus is on predicting sales not on predicting individual user probability of being influenced. We create consumer segments based on attitudinal dimensions that have limited ad targeting value as they do not map well to targetable audiences.
We create attribute profiles of brands but we do not analyze RESPONSE PROFILES. What the heck is that? Each major ad targeting platform (e.g. AOL Platforms, Google, Facebook, Twitter, data aggregators, Datalogix, ShareThis, etc.) offers audiences based on consumer interests and behaviors. How about if research stores campaign conversion rates by these response profiles for each brand?
Hypothetically, consider this. Unilever’s Dove brand team might learn that users who care about beauty, the environment, organic foods, buy Dove 25-67% of the time all respond well to their advertising. They might even find that this pattern transcends age and gender. Think of how different that is from survey attributes that would tell Dove that “brand trust”, and “leaves my skin feeling soft” drive Dove and that the brand indexes high on females aged 35+. The former is a prediction engine that guides advertising action, driving up marketing ROI; the latter is just an insight that quickly gets pretty stale.
Imagine a test and learn, data driven approach.
Turn the first week of any digital campaign into a naturally occurring experiment. Move money around based on what is working. Store results in a database of ad, brand, and user characteristics for reanalysis to improve media strategies for the next effort. Imagine the speed to learning about smart watches if you experiment with ad effectiveness vs. waiting until you have years of data for a regression model!
Consider new product testing now that we are thinking user specific targeting. Sales estimation tools focus on concept testing to determine trial and volume forecasts for a particular new product idea. They make a forecast, then say, “we’ll be back in 12 months to find out how the launch did so we can validate the forecast”. Well, how about a new product forecasting tool that provided targeting information against the same audience profiles I mentioned up above? How about if the forecasting system actually identified a predictive model at a user level to guide impression targeting to those most likely to become triers, using targetable audiences?!
Understanding the relationship of brands to consumer life is a big stated challenge for marketers wanting to be contemporary but these are questions that typically don’t make it into the brand survey…they are left on the cutting room floor in favor of brand equity, ad awareness and attribute ratings. If a marketer wants to learn the relationship of its brand to culture and society, use social media! For example, it is now possible for Verizon to match its customer list to social media profiles and achieve a 30-50% match. Now imagine that Verizon decides to track social media conversation for segments of customers who love Verizon vs. those modeled to be vulnerable to switching. How do smartphones affect their lives? What are their life challenges? What celebrities and TV shows do they talk about? What hashtags do they use? What Twitter audiences are they in? There is richness and hard targeting value in social media data that is not available via surveys.
The research teams at Media companies are better at harnessing their data and developing user-specific media strategies because it is part of their selling narrative but even so, there is an ecosystem problem. Media companies might have rich audiences to offer that, when profiled against poorly constructed consumer segments, show little targeting value. Media companies need to guide advertisers to do a better job of segmenting their consumers, so segmentation projects ultimately result in productivity lift. A new service they can offer to solidify client relationships?
So marketing researchers, make 2015 the year you get serious about data-driven strategies that bring true value and productivity improvement to the marketing function.
How data-driven are you in practice? Start with an objective audit, let’s prioritize classes of research methods to transform, and let’s prove out the data-driven reinvention model I am proposing.