“When digital [advertising] first came about,” said Amanda Martin, Goodway Group VP of Enterprise Partnerships, “the whole mindset was, ‘now we can measure with granularity.’”
“Third-party cookies and [mobile] device IDs created a uniform solution,” Martin continued. It was a sea change in marketers’ thinking about measurement, since ads on TV, radio, and print couldn’t offer that kind of purported exactitude.
Now, as marketers face the demise of third-party cookies and the lesser availability of mobile device IDs, another sea change in measurement is underway.
The new landscape
To better determine the contours of this latest imperfect measurement, let’s scope out what the landscape for ad targeting will probably look like.
User consent is now the key driver. While many marketers and data firms in the early days of online advertising considered audience data to belong to those who acquired and managed it, it’s now considered a core principle that users must consent to the employment of their profile and behavioral data for advertising. This move to a new center of gravity has been primarily driven by browser-makers and privacy laws such as the European Union’s General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) and California Privacy Rights Act (CPRA).
Similarly, marketing measurement now starts with some sort of user consent. Whether marketers are measuring audience reach, sales conversion, or app install rates from specific campaigns, the linchpin is that they are all built around user consent.
With this new consideration, there are five major, co-existing directions for this new era of online advertising—and for the resulting measurement—that are gaining traction.
The five domains
Brands’ islands. With consent, a brand can see when visitors and customers come to its owned properties by using first-party cookies and hashed/encrypted identifiers, such as email addresses. This requires users to log in to the site at some reasonable interval, and creates personalized experiences on owned properties—think of the way a retailer may show certain products to a logged-in user who last purchased a skin-care product, or how a sportswear brand that knows a user is a runner may show the season’s best shoes or an interview with a sponsored marathoner.
Brands’ archipelagos. Individual brands or others connect their silos of user IDs and behavioral patterns in some way, as brands collaborate in an effort to compete with the logged-in targeting available to walled gardens. Brands can then use this enriched data to supplement their existing audiences for targeting across other channels.
People-based marketing. LiveRamp, The Trade Desk, and others use authenticated identities from log-ins, email addresses, offline purchase data, and other sources to anchor deterministic data sets and device graphs of people using first-party and some third-party data. These individuals—not unknown users based on browsing data—can then be reached across devices and brands, leveraging a privacy-forward ID.
Pseudonymous cohorts. Consenting users are assigned to large anonymized groups, such as users whose browsing indicates they’re interested in sports cars, French cuisine, and bowling. These groups are at the core of Google’s FLoC (Federated Learning of Cohorts) proposal. Group membership is stored in the browser and presented anonymously to advertisers. This methodology differs from leveraging deterministic/probabilistic identity, as marketing to the group is based on inferred preferences.
Contextual advertising. This is “back to the good old days,” when advertisers displayed camping equipment ads near camping-related content instead of following a user who had once visited a camping-related page with camping equipment ads across subsequent websites. As with pseudonymized groups, this infers interest and doesn’t rely on identifiers. Depending how close to a campaign a sale or response is activated, a marketer can directly or indirectly measure the impact.
Authenticated, people-based identities
With these five main ways of reaching existing and potential customers, two factors stand out for their increased importance: first-party data and interest-based targeting. As a result, the new measurement is also centered around these two factors.
For data connectivity platform LiveRamp, authenticated and people-based identities are the center ring.
“Our focus is on authenticated addressability with logged-in users,” Travis Clinger, LiveRamp SVP for Addressability and Ecosystem said. “We see measurement shifting to first-party data.”
He pointed out that deterministic identities result in a much higher quality of data for both activation and measurement, as Fitbit discovered when leveraging the LiveRamp Authenticated Identity Infrastructure for a recent campaign. Per the published case study, the brand’s objective was to reach high-value audiences and measure on traditional outcomes without third-party cookies. Users deterministically identified through LiveRamp were then reached on high-quality, authenticated publisher inventory.
That campaign reported a 13% increase in average order value (AOV) vs. third-party cookie targeting, indicating that the audiences reached generated a higher return than targeting on cookies. Additionally, return on ad spend (ROAS) was 2X higher than with cookie targeting. This kind of campaign, Clinger said, is “based on the fact that 100% of their impressions are measurable.”
“When publishers plan for third-party cookie deprecation by focusing on their own authenticated users,” MediaMath Chief Product Officer Anudit Vikram shared, “new insights can be derived through machine learning.”
For instance, he recently ran a campaign with the goal of measuring a million users, but employed 100,000 verified users and machine learning to extrapolate the reactions and behaviors of unknown users, based on such similar conditions as page content or time of day.
As for pseudonymized cohorts, such as Google’s FLoC, he shared that there hasn’t been enough real-world testing to determine how well their targeting or measurement capabilities will actually work. In addition, although user consent is implied for the cohorts, there are also some questions about whether pseudonymized users could be unwillingly identified by cross-referencing the particular cohorts to which they belong.
“Marketers also need to adjust their sights,” said OpenX Chief Architect Joel Meyer, “to understand that consented but pseudonymous users can offer one set of measurements, while consented but identified users provide another set.”
For example, ads might be targeted at 10,000 users in a pseudonymized FLoC group of science fiction fans. Ads for a new sci-fi book might lead to 200 sales among that group, but which 200 users? A purchase may expose a user within a cohort, assuming consent. Some visitors to a brand’s site might consent to having their first-party cookies associated with the pseudonymized sci-fi cohort they agree to join, possibly making it easier to determine that 45 of the customers/visitors in the cohort actually made a purchase among the 200 cohort sales. The brand would then need to extrapolate the attributes of those 45 to the other buyers in the cohort—a task made easier if the data sets are larger.
The value exchange
The first step marketers and publishers need to take in this new post-cookie world, said Index Exchange SVP Product Michael McNeely, is to “get the ‘value trade-off’ going.”
This value trade-off, or exchange, would create a new level of relationship with customers and visitors that encourages them to share more information about themselves and consent to its use, because they understand what they receive in return: discounts, promotions, loyalty programs, exclusive content, and the like. When people actively consent to share more about themselves, brands, publishers, and others collecting first-party data can sharpen their customer intelligence for better targeting and measurement that may uncover insights teams outside marketing would find beneficial.
Focus on getting more authenticated users and first-party data
Targeting and measurement that makes greater use of first-party data and pseudonymized groups based on interests benefit marketers, according to MediaMath’s Vikram. “The more you can use first-party data,” he said, “the better you’ll measure performance on a user basis, and the more FLoC [or similar groups], it’s easier to achieve scale.”
While third-party cookies were useful for marketers, Vikram conceded that “the quality [of data] was up in the air.” He noted that when so much targeting was available via third-party cookies, marketers didn’t put resources into getting more authenticated users and more extensive first-party data, both of which provide much higher quality of measurement.
As this newest sea change evolves, “we will be in choppy waters for a little bit,” he said, but marketers will eventually emerge into “a position with very high-quality data.”