What the Tech is attribution modeling?
Illustration by Ollie Canton / The Current
One of the biggest headaches in digital advertising is the lack of standardization. Different platforms and publishers use different methods and metrics for targeting audiences, as well as measuring campaign performance. This, as you can imagine, has the potential to create difficulties for marketers running cross-platform campaigns.
Nowhere is this problem more pronounced than when it comes to attribution — the process of determining how each component of a campaign contributes to its success.
How a company constructs its attribution measurement system is known as attribution modeling. The conversation of measurement has become even more complex lately with the resurgence of media mix modeling (MMM) — a technique that aims to connect ad exposure to business outcomes, but is distinct from attribution modeling.
In the latest installment of What the Tech, we break down attribution modeling, its differences from MMM and the relative strengths of each.
What is attribution?
Attribution is how brands assign credit to the different elements of a campaign.
Modern ad campaigns tend to span different channels — from social media platforms and display ads on the open web to video ads that run on YouTube or connected TV channels. Each platform often has its own (usually opaque) measurement system, and few are eager to share that data. This makes it difficult for brands to get a holistic measure of campaign performance.
A brand might know how many people saw their CTV ad or liked a social media post, but not how those touchpoints worked together to influence someone to buy. This is where attribution modeling comes in.
What is attribution modeling?
Attribution modeling is a way to solve the tricky attribution problem.
It’s a software-based system that ingests all the data from across a campaign — social, display, CTV, etc. — and determines how much each component contributed to the desired outcome. It then assigns credit accordingly. The result is a performance report that helps marketers identify what worked, what didn’t, and where to adjust ad budget for optimal effect.
Brands, agencies and tech firms have built their own attribution models to reduce ad waste and improve return on ad spend (ROAS).
What metrics does attribution use?
This is where attribution gets contentious.
For years, digital advertising relied on last-click (or last-touch) attribution — giving full credit to the last ad a consumer saw before converting.
Let’s use an example: Say a shoe brand runs a multi-channel campaign, with ads running across display, social and CTV. The potential customer sees the shoe ad on The Washington Post, later spots it on Hulu, and finally sees it on Instagram before making a purchase. Under the last-click attribution rule, only Instagram gets credit for converting that consumer into a buyer. The Washington Post receives no credit, despite serving the ad that initially captured the reader’s attention, and the Hulu ad also ends up with no recognition.
That’s why marketers now use multi-touch attribution, which assigns credit to all relevant touchpoints in the journey.
Are there different kinds of multi-touch attribution?
Excellent question. Yes, there are.
- Linear attribution gives equal credit to every part of a campaign.
- Time-decay attribution gives relatively more credit to behaviors that occur later in the customer journey, closer to the eventual purchase.
There are other models too, each weighing interactions differently depending on campaign goals and customer behavior.
How does attribution modeling differ from MMM?
The simple answer is that MMM offers a broader, more strategic view of campaign performance — but it’s time- and cost-intensive to produce.
MMM adds more data layers to the mix, including sales, regional trends, seasonality and even external market factors. It compares outcomes among people exposed to an ad campaign versus people who weren’t, to estimate whether the ad drove actual business results.
Conducting an MMM analysis can give brands a more precise measure of their ROAS, which is considered the ultimate metric in advertising.
Why not use MMM all the time then?
Because MMMs take a long time to prepare — at least historically.
For decades, an MMM analysis would require months, if not years, of data crunching. But with better tools and faster data access, that’s changing. Some modern MMM tools — powered by AI — claim to produce MMM reports in a matter of hours, or even seconds. Still, MMMs are more appropriate for long-term measurement.
Attribution modeling remains essential for real-time, tactical decisions, especially for campaigns with shorter-term goals, such as generating reach, engagement, email sign-ups or general brand awareness.
Think of attribution modeling as the tactical compass that helps to navigate the day-to-day, while MMM reveals the big picture. In today’s media world, both speed and strategy matter — and measurement should deliver on both.