The Brand Marketer'sGuide to Attribution Models in Commerce Media

‍

Attribution sounds like a technical detail. Something for the analytics team to sort out while the brand team focuses on the fun stuff.

It isn't. Attribution is the set of rules that decides which ad gets credit for a sale. And whatever gets credit gets budget. Change the rules, and you change where millions of dollars go, often without anyone noticing.

If you're a brand marketer running commerce media, you don't need to become a statistician. But you do need to understand how the scoring works, because it's shaping your plan whether you're paying attention or not.

‍

The basic question attribution answers

‍

A shopper sees a display ad on a retailer's site Monday, a sponsored product Wednesday, an offsite video ad Thursday, and buys your product Saturday. Which one gets credit?

Different attribution models give different answers. None of them is "the truth." Each is a rule for splitting credit, and each has blind spots.

‍

The common models, in plain English

‍

Last-touch. The last ad the shopper interacted with before buying gets all the credit. Simple and common. It tends to favor bottom-of-funnel tactics like search and sponsored products, and it undervalues the ads that built interest earlier.

First-touch. The first ad gets all the credit. Rare in commerce media, but it shows which tactics start journeys. It ignores everything that closed the deal.

Linear. Every touchpoint shares credit equally. Fair-sounding, but it assumes every ad mattered the same amount, which is rarely true.

Time decay. Touchpoints closer to the purchase get more credit. A reasonable middle ground for shorter buying cycles.

Position-based. The first and last touches get most of the credit, and the middle ones share the rest. Tries to honor both discovery and conversion.

Data-driven or algorithmic. A model uses actual conversion data to estimate how much each touchpoint contributes. Potentially the most accurate, but it's often a black box, and it still only sees what that platform can see.

‍

The settings that matter as much as the model

‍

Even within one model, a few settings dramatically change the numbers.

Attribution window. How long after an ad a purchase still counts. A longer window captures more sales, but it also gives credit for purchases that might have happened anyway. Windows vary across networks, so make sure you know what each one uses before you compare them.

Click vs. view. Does a shopper need to click the ad, or just see it? View-through attribution can be meaningful for awareness formats, but it can also inflate results, especially when ads are shown to people who were already going to buy.

Brand halo. Some networks count sales of any product in your brand, not just the one advertised. That can be fair (ads build the brand) or generous (the shopper bought your other product for unrelated reasons). Know which is being counted.

Online vs. in-store. Some networks can attribute in-store purchases through loyalty data, others only online. That changes the totals and makes networks hard to compare.

‍

Attribution's big limitation

‍

All attribution models share one blind spot: they credit ads for sales, but they can't tell whether the sale would have happened without the ad. That's the job of incrementality testing.

Think of it like this. Attribution is the scoreboard each team keeps for itself. Incrementality testing is the independent referee. You need the scoreboard to run the game day to day. You need the referee to know who actually won.

‍

How to use attribution without being fooled by it

‍

Know the rules for every network. Make a simple cheat sheet: model, window, click or view, brand halo, online or in-store. Keep it updated.

Compare like with like. Before comparing ROAS across networks, normalize as much as possible, or at least note the differences on the slide.

Use attribution for optimization, not big strategic calls. It's great for deciding which keyword or creative to push within a campaign. It's weaker for deciding how to split budget across networks or channels.

Calibrate with lift tests. Run incrementality studies periodically and use the results to adjust how much you trust each network's attribution.

Watch for funnel bias. If your plan keeps drifting toward bottom-funnel tactics because they "perform best," check whether your attribution model is quietly rewarding them for catching shoppers who were already on their way.

‍

A quick scenario

‍

A beverage brand compares two retail media networks. One shows much higher ROAS than the other, so the team is ready to shift budget.

Then someone checks the fine print. The higher-ROAS network uses a longer attribution window, counts view-throughs and includes brand halo sales. The other uses a shorter, click-only window on promoted products only. On a normalized basis, the gap mostly disappears. A follow-up lift test shows the "weaker" network is actually driving more new buyers. The budget shift doesn't happen, and the team adds an attribution cheat sheet to every quarterly review.

‍

The bottom line

‍

Attribution is a rulebook, not a reality check. Know the rules each network plays by, compare carefully, use attribution for day-to-day tuning and lean on incrementality for the big calls.

Five Eighty helps brands make sense of attribution across retail media networks, so budget follows real impact instead of whoever keeps the most generous scoreboard. If your ROAS comparisons feel too good to be true, let's check the rulebook together.