“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Mark Sturino, VP of data and analytics, at Good Apple.
The “death of the cookie” provides an opportunity for all brands to learn from pharmaceutical marketers and leverage new AI and statistics software, creating models that solve some of the biggest tracking problems – cookies, cross device tracking, and walled gardens.
Cookie replacement solutions connecting first-party data to individual ads through universal IDs are coming, but rather than chasing a retooled version of a historically clunky solution, marketers should build new data frameworks that employ statistical modeling and AI to illustrate a probabilistic media journey.
The pharmaceutical industry can serve as a point of reference for this new framework – though it typically lags years behind other industries in technological adoption due to its heavy regulation. These companies have had to work within privacy-compliant guidelines for years now.
Strict privacy laws long ago stymied the cookie for simple measurement purposes, leaving marketers to turn to third-party, HIPAA-compliant companies that provided effective, though expensive, solutions to tie media exposures to prescriptions and measure ROI.
Ahead-of-the-curve pharma marketers knew how to use that data when it was available. They also spent time looking at old-school mixed media models and added technology and a ton of data to build out demographically, geographically and behaviorally segmented modeling processes to solve for ROI. This strategy can be combined with third-party pathing results to effectively optimize media tactics and mixes. Integrating these models into reporting allows the whole system to be automated and used on a weekly basis for optimization.
The specifics of how this can be done will vary by industry and primary KPIs; building plans that consider media mixes at the DMA level with purposeful changes in budgets, channel mix, and messaging in previously identified markets will provide data scientists the necessary variance in data across similar markets to run better analyses.
Think of these analyses as elaborate, ongoing matched-market testing in which audiences are segmented by DMA and then similar segments are exposed to different media plans to allow probabilistic modeling. This helps media teams better differentiate what is actually driving success. Planned changes by DMA can help test new media, updated messaging, etc.
Say you’re a CPG marketer. By integrating a data scientist with expertise in these types of media mixes into your media planning process, you will better optimize and measure against your specific business objectives. However, as part of this process you will also have to make media planning decisions, for the sake of measurement, that may go against best practices in media planning.
As a simplified example, where you have two similar DMAs, plan one DMA as a test market to allow for the impact differences to appear in order to better understand the causality of the media. This strategy may go against most brand marketers’ instincts for how to handle similar audiences, but it will give you the information you need to better engage with the entire segment of like-minded shoppers going forward.
An additional advantage of a probabilistic approach looking at data from population segments rather than individuals is that it can be more directly tied to key ROI-driving tactics. Cookie tracking prioritized proxy metrics that are easy to obtain over actual outcomes that matter. The loss of cookies will force all marketers to become more strategic about how we use data, just as it has already forced pharmaceutical marketers to continue to evolve an older measurement methodology. A better connection of media data to sales will allow for better ROI conversations with brand clients, and connecting marketing efforts to ROI is clearly fundamental to success.
As seen in pharmaceutical marketing, this new way of working also provides brand reputation benefits by giving the cautious marketer more incentive to pay higher upfront costs for quality inventory, ad monitoring and fraud detection because in the end, ROI, not CPMs, will drive the decisions. A family-friendly detergent brand doesn’t want inventory next to upsetting political material. Poor quality ads and ad environments may be great at driving impressions and traffic, but they do not lead to purchases and can damage brands. Putting more effort and budget into quality data mixes helps build a reputation while still optimizing a campaign.
This may be a major structural shift for many brands, but the tracking methods commonly used were due for change. This evolution creates an opportunity to reconsider the data problem as a whole, and explore new paths for measurement success. Deterministic views will play a role, but that should not be the only solution.
While building out data flows and integrating new technologies will not be easy, it won’t be any more difficult than building out the deterministic path the industry is currently looking to move towards.
Follow AdExchanger (@adexchanger) on Twitter.