Approaching pillars of the marketing loop in new ways
Spot the difference?
We can’t either. Targeting, delivery, measurement, and optimization and/or modeling remain the pillars of the traditional marketing loop. While they’re here to stay, the ways we approach them are changing in big ways.
You’re bound to be familiar with some of the reasons why, but let’s unpack them.
Change is starting from platform changes. Under the banner of user privacy, every operator of a web browser has made or announced significant changes to data collection that will make it much harder to collect behavioral data on individual users.
Effective sometime between March and September 2022, cookies will no longer be usable as a targetable identifier in DV360, and Google is walking away from 1:1 targeting altogether by removing support for UID 2.0 and other identity solutions from DV360. Meanwhile, Apple is making IDFA opt-in on its devices, severely limiting targeting capabilities on mobile.
Since Google, Apple and others are just now catching up to where consumers are in terms of privacy expectations, it follows that the new marketing loop will be much more privacy-focused. To visualize this, I like to imagine the double helix, which you’re probably familiar with from images of DNA.
Think about privacy and identity management being intertwined with technology systems where consent is built in by design, and that’s the DNA of the future advertising ecosystem with the consumer at the forefront.
The marketing loop
Next, let’s explore each section of the marketing loop in turn to see what the coming changes may bring to the future of advertising.
The demise of third-party cookies will lead to greater reliance on first-party data and second-party data collected by the “triopoly” (i.e., Google, Amazon and Facebook), as well as deeper partnerships with major publishers who have their own authenticated traffic and datasets.
It’s important to remember, though, that not every brand has a wealth of first-party data. (Think of CPG brands, which typically sell through retailers and don’t own the relationship with their customers.)
It’s clear that we’re moving from a paradigm of 1:1 targeting to aggregate, cohort-based models in some environments.
This will ultimately enable more rigorous marketing science by giving data scientists access to different types of datasets to query and analyze. Instead of relying on cookie-based tracking to provide a pipeline of behavioral data about individual users, we’ll focus on group behaviors, pair that data with contextual or intent signals and then run it through machine learning models to achieve scale.
Additionally, if a brand has access to some PII data, we may help the brand achieve scale by uploading its list of 10,000 customer emails to a walled garden clean room and build look-alike or act-alike models to activate within that environment. Or we might upload that email list directly to a DSP in order to leverage its identity solution (like The Trade Desk’s UID 2.0) to scale audiences.
The relevant “groups” or “cohorts” for 2023 and beyond will comprise between 1,000 and 10,000 individuals and be machine-defined based on patterns, and consumers will belong to only one cohort at a time.
Google is currently advocating for Federated Learning of Cohorts (or FLoC) to be the privacy-compliant mechanism for ad targeting that replaces cookie tracking, but it’s unclear whether the industry will rally around it. FLoC doesn’t allow for sharing of users’ browsing history but instead creates interest-based cohorts, such as audiences that are likely to convert on fitness and healthy lifestyle, which are refreshed on a weekly basis.
FLoC is promising in many ways but still fails to address the issue of user-level consent management and may introduce bias into audience creation (if, for example, people are grouped into sought-after audiences, such as in-market SUV buyers, based on weak signals). Mozilla, Microsoft Edge, Brave and Vivaldi are all currently declining to use it.
The bottom line is that how cohort-based targeting will evolve is a big unknown, and its efficacy is yet to be tested later this year.
The jury is out on how delivery will change in 2023 and beyond. What’s clear is that the triopoly and premium publishers like the New York Times, Condé Nast and Meredith, with large, often logged-in user bases, will gain a bigger share of the ad pie on the strength of the data they collect about their users.
That said, many brands have started to see diminishing returns on Facebook, and each fresh controversy and scandal on the platform is creating an aversion on ethical grounds.
I expect to see growing numbers of marketers conduct rigorous audits of where their ad dollars are flowing to make sure those platforms and publishers align with their brand values.
Meanwhile, programmatic advertising on the open web is going to need to develop new models to stay competitive. There will be cohort-based approaches (e.g., FLoC), as well as attribute-based fusion that leverages machine learning and other statistical techniques to provide individual-level precision matches—all without identity joins. 1:1 targeting will continue to be possible through the likes of ATS and UID 2.0 consortia targeting tied to hashed PII identifiers, such as email addresses.
We have yet to see how scalable or addressable these consortia-developed solutions will be versus cohort-based targeting. Some experts speculate only 20% of the open web will have authenticated traffic, so the most important question remains:
Will advertisers be willing to spend on sites with non-authenticated audiences, and how will their efficacy and reach differ from their traditional KPIs?
Future-state measurement solutions are quickly evolving in response to the impending changes. At Hearts, we expect to maintain support for third-party measurement solutions and 1:1 attribution where relevant, but we also expect lift analysis, extrapolated (ID-less) measurement solutions, and browser and environment-specific solutions (e.g., Privacy Sandbox and SKAdNetwork) to give advertisers more insight into performance.
It’s clear that measurement will have to leverage multiple solutions to make up for the loss of cookie tracking and the ability to deconstruct customer journeys (albeit in tiny sample sizes).
The fact that users won’t always be logged in is what makes it tricky since our view of what they do online will be severely curtailed. This means we’ll have to use deterministic methods as a baseline and then leverage probabilistic modeling to understand audiences on a larger scale.
Some good news is that “clean rooms” set up by the triopoly, as well as by platforms like LinkedIn, Snap and Verizon to comply with privacy regulation, can provide much deeper insight into performance.
Instead of receiving data on individual customer journeys like we used to, the aggregate data we now get from clean rooms has much more scale, making it more accurate for modeling. And since we can’t take any of it out of the clean room environment, some platforms are also granting access to more “walled garden” data than they used to.
For example, we can explore logged-in activity on YouTube, tie it to Google affinity audiences in a privacy-safe environment and add DSP or other partner data to the mix, which is incredibly significant since we never had access to log-level YouTube insights before.
New types of measurement will also arise as the user journey continues its collapse across third party cookie dependent environments. Lost signals, such as view-through conversion visibility going away, will push the industry to place a greater emphasis on quality and attention as the new currencies of media measurement.
In a cookie-less world, optimizing against conversion will be much harder, and it won’t occur in real time due to a lag in browser and ad server syncing.
There will be multiple conversion measurement challenges for modeling, such as device linking when multiple devices are used in the customer journey and new privacy regulation.
The industry will need to embrace machine learning-driven attribution approaches that account for time-series connectivity, such as the ones now deployed to improve traditional Media Mix Modeling (MMM). We can also leverage a combination of MMM and multi-touch attribution (MTA) techniques to start optimizing campaigns in clean rooms in real time.
Models like MMM will be used to normalize results across clean rooms environments and to scale overall channel performance in MTA. Welcome back to Markov chain stochastic models everyone.
How is Hearts & Science preparing?
To get ready for whatever the future may hold, we at Hearts are acting as Agents of Change, investing in five key areas:
- Assessing the use of first-party data in open web activation and building alternative options. (See my colleague Sarah Polli’s great piece on how marketers can mix and match their first-party data with second-party data and contextual data.)
- Developing a hybrid clean room strategy to join walled garden environments.
- Investing in cutting-edge techniques for probabilistic modeling.
- Building robust A/B testing across all media to understand media impact.
- Exploring new metrics and measurement approaches in a post-cookie world. (See JoAnn Sciarrino’s piece on the growing importance of “brand attachment” and our Marketing Science Team’s piece on why it’s time to focus on quality and attention.)
We’re still effectively in the first inning of a very long game, and many factors will have a bearing on what the marketing loop looks like in 2025 or 2030. We have yet to see whether Alphabet will include Android device IDs in its 2022 identity deprecation, and Microsoft hasn’t announced any moves with Windows or its browsers. And that’s only scratching the surface of all the unknowns.
Change is unavoidably scary, but we at Hearts are looking at this moment as an opportunity to build the marketing loop back stronger than before, making it more rigorous and data-driven while honoring user privacy. Our new Agents of Change series will look at transformation across three key areas: analytics, data and technology and media measurement in a new privacy-first world. We’ll unpack how the game is changing in each of these domains and the opportunity these changes create for brands.