The digital ad sphere is about to change drastically. With privacy and user protection in the spotlight, Apple and Google are both in the process of phasing out third-party cookies. Digital media has long been centered around that third-party cookie. That little file tracks your exposure, interaction, and conversion and has been the cornerstone on which media is targeted and tracked, allowing advertisers to identify users across platforms and channels. Losing this tracking ability leaves advertisers in the dark on how their external ads are contributing to conversions.

When these cookies are finally blocked, the current sphere of digital advertising analytics will be forever changed. However, we can use the external ad tracking data we currently have to build predictive models and make informed decisions for years to come. In this article, we’ll explore how ad tracking analytics works now, what will be changing in the near future, and how we can use current ad tracking models to benefit our future advertising endeavors.

How Digital Advertising Analytics Currently Work

Modern digital marketing analytics is both messy and intricate. With many users often being exposed to many types of display and video ads across many platforms, the relationship between consumer and advertiser is becoming increasingly complex. In these complex relationships with many different touchpoints, it is difficult to determine which interactions and channels are effective, leading toward conversions, and which are ineffective, wasting precious advertising spend or driving the consumer away from conversion.

In order to understand the impact of these customer media journeys, multi-touch attribution models have become the industry gold standard. MTA models are marketing analytics tools that help determine the impact each touchpoint has with a consumer on their path towards conversion. These methods allow us to assign value to marketing activities, e.g. native ad impressions, promotional emails, clicks, page views, within an analytics or customer data platform. This then allows us to understand how each interaction drives conversions and informs overall marketing strategy.

One of the keys to an MTA model is granular, log-level data which allows us to see how a specific user interacted with our ad, website, and/or video. To enable this low-level data, we need external cookies that allow us to identify a user between platforms and channels; it is important to know that one user did all of these activities in order to correctly weigh the ad’s effect. These user interactions allow us to analyze each user’s marketing journey.

The way these paths are constructed changes depending on whether you use a simpler MTA model, like a linear attribution where each interaction has the same weight, or first/last touch model where the first/last touch is given the most attribution. Other algorithmic approaches weigh each ad less with time/position.

More advanced models, like Markov Chains or Shapley Analysis, allow us to more accurately identify how interactions influence conversion. With Markov Chains, we build a delimited path, e.g., ”, then use these paths to identify the probability that a user will move from one step to another, and then compare those probabilities when those events are removed. This allows us to determine the value of these activities; for example, we might find that users will easily convert after viewing an email ad, given that the user has had at least one banner view in the past.

These findings allow us to target more specifically, identify valuable channels and interactions, and better understand how users are influenced by advertisements.

Media Optimization and the Evolution of Privacy

As previously mentioned, the low-level data around users’ activities (impressions, clicks, etc.) is key to building these models. While user log data is available now, new privacy concerns are going to be limiting the availability and breadth of these log data. To protect customer data, Apple has already blocked third-party cookies for Safari, and Mozilla has blocked them for Firefox as well. Following Safari and Firefox’s lead, Google Chrome plans to phase out external ad tracking by 2022.

Not only will the majority of web browsers stop supporting third-party cookies, but mobile phone activity tracking will be similarly limited. During Apple’s WWDC 2020 Keynote event, iOS 14, due to be released in the coming months, was announced to include a new dialogue presented to each user with each app they open:

Example of warning

This means if people don’t choose to allow tracking (and why would they?), and there is no tracking on web browsers, then we lose a large amount of data with the ability to tie impressions to clicks and/or conversions, in short, losing the insight into how users are interacting with external advertisements.  Advertisers will be limited to impressions and clicks, while losing insight into reach, frequency and conversion. This will completely disrupt current marketing data strategies.

Although most browsers will be blocking third-party cookies, brands are still able to use first-party cookies to measure site-side activities. By using the current tracking data available, coupled with first-party cookie data that is available indefinitely, brands will be able to prepare themselves for the loss of third-party cookies. Brands must manage two aspects of their media strategy: focusing attribution on site-side activities as opposed to media interactions and optimizing digital media in line with traditional media.

If you planned to use this external third-party data as a part of your long-term strategy, now is the time to create the MTA models and glean insights while the data is still available.

How Can We Use These Models After We Lose Third-party Cookies?

The attribution process slices up the value of the user’s visit and estimates the extent to which each site-side interaction in the user’s journey contributed toward a sale.

In general, this type of consumer behavioral data can be a gold mine for marketers and advertisers. Through their behavior, consumers are constantly providing brands with feedback and a truer measure of their intentions than can be gathered from surveys. In this mindset, one can view each consumer’s actions as an indication of purchase intent that can be used to quantify the value of a visit. Once the dollar value of each visit is known, that information can serve as the basis for attribution value. Simply put, site-side MTA evaluates the impact of each aspect of the brand’s website on the likelihood of purchasing.

Without third-party cookies, the only media that can be tracked will be the last media clicked. Custom landing pages will enable first-party site-side cookies to know which media drove the visit. If this click leads directly to a sale on the site then the sale attribution is clear.  Any other media in the consumers path is un-trackable.  With these constraints any awareness-based media is likely to show little to no value.

However, with site-side MTA each activity that precedes a sale is valued.  If a consumer compares product details, reads reviews or performs a similar activity without a sale, the site-side attribution can provide a revenue value from each of those upper funnel activities.  This allows awareness media, which may not be focused on sales conversion, to show direct value. Now the site-side consumer journey rather than the media-side consumer journey is what drives attribution and media reporting.

Going further, once the value per website activity is known then tracking media impact becomes a straightforward process. At this point, all media clicks going to the site can receive a dollar value per visit based on the count of the high-value activities identified, weighted by their sales contribution. When combined with media cost, the value delivered can create an attributed ROI by channel. This formulation allows the trackability of web activities to optimize media since it is no longer focused only on a sale, but prior site visits that help build toward that sale are attributed a portion of that value.

Digital Media Optimization Without Cookies

As the industry shifts away from trackable cookies, the impact of media execution on sales can be captured by a media spend based model. These models enable not only a historic analysis, but also utilize the discovered relationships to produce ongoing reporting, ROI, and optimization opportunities.

With this formulation, site-side attributed ROI showcases the amount of incremental profit received for each dollar spent. Media data are examined at the most granular level available, the key principle being that media are estimated at the level at which tactical decisions can be made. This may be at the channel level, provider level or occasionally the placement level. The goal is to balance the robustness of the model with the optimal level of granularity.

This becomes important not only in reporting but also in optimization efforts. As all clicked media can now be tracked in terms of the value delivered and the attributed ROI, the derived value and media spend can be integrated into logistic curves to determine optimal thresholds by channel. The true power of these models are the diminishing returns curves and the concept of marginal return. Once the concept of marginal returns is realized, the last dollar spent does not have the same value as either the first dollar or the middle dollar. This allows for a finer understanding of where to invest in media across channels.

In a world without third-party cookies, first-party cookies become king. With extensive site-side tracking, we will still be able to understand media performance and optimization. Although external ad tracking may be no more, multi-touch attribution is evolving to no longer need them.

Author: Jonathan Prantner

Teacher. Innovator. Wizard behind the curtain. Jonathan’s approach to applied mathematics has pushed analytics to the limits for over 2 decades. A celebrated thought-leader and recipient of multiple data science patents, Jonathan Prantner is currently the Chief Analytics Officer and Co-founder of RXA.