BackHow mitigating identity impacts multi-touch attribution
Accurate and reliable measurement is at the heart of determining media effectiveness. Matching media activity to online and offline conversion activity relies on accurate, reliable identity matching.
The changes imposed by Apple through IDFA and the impending deprecation of third-party cookies on Chrome (now pushed back to late 2023) threaten to have a significant impact on multi-touch attribution (MTA) and last-touch attribution (LTA). These are used to measure billions of dollars in ad spend.
While there are several measurement techniques available to marketers, this piece will focus on identity impacts to MTA and provide an overview of where MTA fits in a comprehensive measurement framework.
First, let’s zoom in on the measurement approaches that predominate today.
The measurement landscape
There are four major measurement types used by most marketing organizations: media mix modeling (MMM), multi-touch attribution (MTA), last-touch attribution (LTA) and site analytics.
Below is a table explaining the role of each — and the expected identity impact caused by IDFA and cookie deprecation.
MMM is the standard for analyzing marketing impact on sales, but it typically requires two to three years of data. It often also lacks the required level of detail for making tactical optimizations.
LTA became an important tool for marketers with the rise of digital media and can be optimized in real time, but since it gives 100% of sales credit to last-touch events, like search, and only tracks online conversions, its results are skewed.
MTA has gained popularity as a method to complement MMM and solve for its shortcomings (i.e., lack of detail and speed to insights). While MMM is intended as a budget planning tool, MTA is more of an optimization and test-and-learn tool. It can be seen as a more mature solution than LTA, providing a bottom-up, tactical layer to complement MMM.
The combined MMM and MTA solution is useful for channel-level planning and within channel optimizations.
Each approach serves a particular purpose and should be used in conjunction with one another.
How to manage identity impacts to MTA
While marketers will not be able to avoid identity impacts to MTA, a well-constructed plan can help them understand where the impact is occurring and how to maintain the accuracy of MTA models to the greatest extent possible. Here are four tips on how to move forward:
1. Collect baseline data now
For each integration or partner your MTA provider is currently measuring, check match rates and how they have trended over time (i.e., the percentage of instances when a user can be tracked across channels).
Any decreases in match rate may indicate an impact from identity matching for a particular integration or partner. Further exploration would be required to understand drivers of the decline.
As match rates vary across partners, devices and browsers over time, it’s important to understand how partners normalize results to ensure their outputs are robust and accurate.
Since the identity graph is the cornerstone of any MTA model, find out which one your MTA provider uses, collect baseline demos and other pertinent characteristics and then look for changes over time.
If media events start to be matched to different types of individuals, it could be a sign that the results of the model are less accurate than they used to be.
It remains to be seen which identity graph solutions will gain the most adoption within the industry, so be sure to ask about your MTA provider’s plan going forward. Do they intend to use The Trade Desk UID 2.0, LiveRamp, Neustar Fabrick or another identity spine? If your provider is not using a widely adopted solution, the accuracy of their models could be impacted.
2. Pursue as many direct partner integrations as possible
Publishers that require users to be logged in can provide what’s known as second-party data. This can be leveraged by advertisers to improve identity match rates and reduce reliance on third-party cookies through deterministic or probabilistic methods against the MTA provider identity graph.
This second-party data can be anything from email addresses to IP addresses to other billing details associated with user accounts.
3. Innovate inside walled gardens
New approaches are being developed that use MMM to combine MTA results between walled gardens like Amazon, Facebook and Google.
Specifically, MTA can be used to determine the attribution of media touches within the walled garden, while MMM is used to scale the overall impact from each walled garden’s MTA results.
4. Explore regression-based attribution
Cloud computing has unlocked the ability to quickly process enormous amounts of granular data. Whereas traditional MMM models typically rely on two to three years of weekly DMA-level data, new genetic algorithms are in development that utilize 12 to 24 months of granular daily digital data.
Model results can include dimensions that typically aren’t possible with MMM but are common in MTA and LTA models, such as publisher/partner, media format and audience. Such granular results provide marketers with much of the detail they seek from MTA without the reliance on cookies or other personal identifiers.
But since this approach utilizes aggregate data, not individual-level data, it comes with one big trade-off: Pathway analysis isn’t possible. In other words, we can’t know that 80% of users who converted through search were potentially influenced by social ads.
MTA isn’t going anywhere
Media effectiveness will always be dependent on measurement in some form. MTA results will continue to be benchmarks for performance after the demise of third-party cookies. The practice of MTA is evolving with the industry — particularly in the area of walled gardens and neutral clean rooms, which are finding creative solutions for identity challenges.
These marketplace changes have also helped highlight the role of MMM and the overall importance of leveraging MMM as a fundamental part of measurement strategy. The good news is that marketers can still rely on MTA to make data-driven decisions about optimizing their marketing budgets — so long as they have a plan for monitoring and managing identity.
