AI and Cross-Channel Attribution: What Marketers Need to Know in a Cookieless World

By Scott Smigler, Division President and Founder at Agital

In 2024, continued advances in artificial intelligence are revolutionizing the ways in which advertisers find, engage, understand, and measure their audiences and overall marketing performance. Quite frankly, the timing couldn’t be better. With the actualization of third-party cookie deprecation underway, “business as usual” protocols for attribution and cross-channel optimization are being completely upended.

AI and machine learning (ML) will help advertisers bridge the cookie gap to reunite (and improve) their understanding of cross-channel campaign outcomes. However, properly tapping into the potential of these tools will require marketers to implement sturdy data infrastructures, define clear goals, and use the appropriate advanced statistical models. Here’s what marketers need to know about the changes that are underway:

Losing Audience Connections to Find New Ones

As the digital marketing landscape transitions to a cookieless environment, marketers should not rely solely on the multi-touch attribution being provided by individual channel partners like Google and Meta. That’s because individual channel advertising AI will find it increasingly difficult to obtain enough observed data to make statistically relevant (and accurate) decisions.

The AI tools of ad platforms won’t become worthless; they give marketers an understanding of what the individual ad channel can see—but they certainly don’t give the full picture.

When using individual channel advertising AI attribution tools, marketers—especially retail brands—run the risk of modeling campaigns based on synthetic data. Depending on a business’ size, these tools might incorrectly determine which campaigns are valuable due to a small sample size or a lack of understanding of relevant, outside factors (like seasonality). Individual channel attribution could also struggle to properly weight high-value products due to their longer purchase cycles, which could cause algorithms to optimize around items with higher purchase frequency but lower prices.

The data needed to build strong cross-channel attribution modeling won’t be available to digital advertising companies like Google, Microsoft, and Meta, but it will be available to the brands and retailers responsible for the data. In order to assess and optimize the impact of their holistic campaign efforts, brands will need to either hire people internally to build new cross-attribution models or share their data with third-party resources or agencies that can build first-party and relational attribution modeling.

Rebuilding Cross-Channel Attribution

Applying data discipline and organizing data under one umbrella will allow marketers to better understand how their campaigns are performing, and which areas need to be optimized. To achieve this, it will be critical for brands to establish relationships with agencies and partners that can help provide and implement the right tools. A tentpole data tool via a platform-agnostic partner can generate insights that ultimately connect what happens at the channel level to profits and losses.

Some of the most useful tools of the future have already been around for some time. The actualization of cookie deprecation underscores the renewed significance of conventional methods and tools like media mixed modeling (MMM) and statistical modeling. MMM is a relational, statistical modeling tool that is, by nature, cookieless. MMM is helpful for all types of businesses when it comes to guiding budgeting decisions across many marketing channels.

Moving forward, brands and retailers will need strong cross-channel attribution modeling tools to compete. In terms of selecting the tools, it’s not necessarily about choosing one over the other; it’s about choosing the right tool for the job based on goals and objectives. Choosing a strong partner that leverages AI, ML, and supervised learning (SL) to innovate is key, but it’s also useful for brands and marketers to have baseline knowledge of the options. For example:

  • Supervised learning (SL) and machine learning (ML) are two categories of algorithms. Both process large quantities of data to identify data patterns through inference, but SL goes a step further to understand data as it relates to a business’s specific needs.
  • Unsupervised ML models require a lot of data. They can help identify potential marketing segments or audiences and recommend which ads to display for different segments. Much of this takes place in the advertising platforms themselves, but it’s important to have a general understanding of how it works. Brands that decide to use these techniques to create their own customer or audience insights will likely need humans to help interpret the clusters and explain how to take full advantage of the insights.
  • Statistical models and ML are both valuable in creating cross-channel attribution models. ML models cost less in manpower but more in processing, while statistical models cost more in manpower but less in processing. ML can create hundreds (or thousands) of models and pick which one is the most accurate in line with a given goal, while statistical models will be tuned around relevant, nuanced business details to answer highly specific and nuanced questions.
  • ML will likely become less expensive as AI tools improve, which will likely cause it to be adopted over manually tuned and more nuanced statistical models. For this reason, we predict ML modeling will become popular for SMBs, while both ML and statistical models will be used by larger enterprises (companies with more money to spend on nuanced, relevant modeling that requires more human time).
  • First-party data attribution tools can also be used to help brands get an understanding of their online campaign performance. These tools typically require some setup and maintenance, and brands need to be clear on their user journey (which can be hard to gauge due to ongoing cookie degradation and non-digital touchpoints like catalogs or store visits). However, for online-only businesses whose primary drivers of performance are paid sources, first-party data attribution tools can be helpful for optimizing campaign budgets and learning from new campaign-level strategies.

Ultimately, the right cross-channel attribution models, organized and supported by the right partner, can yield actionable insights that connect channel-level activities to the bottom line. Today, that can be a competitive advantage for a smart business, but tomorrow, it will be table stakes for remaining relevant.