2024 was a year of hectic change for ad measurement.

Third-party cookies may have been given a reprieve by Chrome, perhaps even an indefinite lifeline. But, still, user-level data is running dry, to the point that last-click and multitouch attribution have lost their edge entirely.

In their place, data-driven advertisers have reverted to familiar stalwarts like media mix modeling (MMM), as well as new techniques like incrementality testing.

These are a few of the new ad attribution trends and techniques programmatic advertiser should know going into 2025.

Incrementality measurement

Media buyers were consumed by “curation” mania this year. But for ad measurement, 2024 was the breakout year of “incrementality,” a hard term to define.

“I really need a better answer to this question,” Olivia Kory, head of strategy at the incrementality measurement startup Haus in an AdExchanger Talks podcast this month, when  asked what actually is incrementality measurement.

She sums it up as a marketing measurement model that is geared toward establishing causation, rather than correlation.

Incrementality measurement achieves this through the sophisticated use of holdout groups and geo-testing. One way to benchmark Instagram’s incremental contribution, say, or that of a large DOOH campaign, is to run that campaign in some markets while not serving those ads at all in other, similar markets.

Early this year, the shoe retailer ASICS detailed to AdExchanger how it uses Habu and the Google data clean room, Ads Data Hub, to create lookalike cities to test for incremental ROI.

The advantage of geo-testing is that it evaluates large cohorts of people (the city of Chicago, say), rather than tracking known users for conversions or changes over time, said Habu’s head of customer success Avanti Gade. This makes incrementality measurement a durable strategy, whereas most user-level attribution vanishes with new privacy standards.

But there are downsides. For one, large platforms don’t necessarily enable incrementality testing. Meta and Google do, but Amazon doesn’t have a ready way for advertisers to test in one city while creating a holdout group of another.

It is also expensive: there are the vendors, the costs of testing and the costs of holdouts.

As ASICS global marketing manager Devin McGuire said, New York City and Los Angeles are blocked from incrementality testing because those cities are too important to the bottom line. Can’t muck the core business up with some advertising test.

But the costs could be negated if incrementality testing reveals a company’s entire branded search budget can be better spent elsewhere.

There’s only one way to find out, and that’s to test the idea.

Mix modeling

Call it vintage chic, because MMM is back.

Not so long ago, data-driven marketers would have scoffed at the idea of a reversion to MMM. It’s an old-school method of campaign measurement built for TV, radio and print, and which takes months to establish results.

But with user-level data running dry and walled gardens hoovering up all the ad demand, MMM becomes a feasible way to attribute platforms as a whole, without having visibility into the platform itself.

The platforms have heard the MMM requests, and answered in 2024.

Google this year launched Meridian, its open-source MMM service. Meta already had one, called Robyn.

The next step for the category is to use data science and new first-party data tools to collapse the timeline for MMM from months to weeks, and hopefully even days, so that results might be incorporated into real-time campaigns.

Another challenge will be to open MMM to more advertisers. It requires large budgets and data analysts, so has been the remit of large TV-heavy advertisers. Now, ad tech vendors are trying to bring MMM to smaller digital or regional brands.

The ecommerce metric plethora

The rise of digital-native brands and ecommerce shopping overall has created new ad measurement metrics that are likely to catch on with other advertisers, especially in retail media.

One biggie is ACOS (“advertising as a cost of sale”), which began as a way of attributing ads on Amazon.

The mathematical difference between ACOS and ROAS is negligible. However, ACOS requires a direct connection to purchase data, because revenue generated by paid media is part of the equation.

But the main difference between ACOS and ROAS is philosophical. ACOS considers ad spend as a contribution to overall sales. For ROAS measurement, the ads are the whole point.

And other boutique ecommerce metrics could spring up.

At AdExchanger’s Programmatic IO in Las Vegas this year, Home Depot head of marketing measurement Zach Darkow said he wished an internal metric the retailer dubs “return on marketing objectives (or ROMO)” would catch on.

Whereas ROAS measurement is a way to justify and expand a specific ad budget, ROMO expands the measurement lens to incorporate factors like brand awareness or category share in stores. He also cited WISS, or “web-influenced store sales,” a short-lived metric at Target Roundel.

Those informal metrics wouldn’t work with walled gardens that only self-report ROAS, Darkow acknowledged.

The common thread throughout the development of new attribution models and metrics is the challenge in educating and inspiring brand marketers to actually change their behavior, as one retail marketer told AdExchanger, when built-in platform ROAS is “sitting there like a feather bed.”