By Nick Perkins, Head of Supplier Analytics, Nectar360
Walk into a supermarket, see a bold feature display and pick up the advertised product. In a last-touch world, where the final interaction gets all the credit, that single in-store prompt wins. It’s tidy, it’s comforting… but it’s often wrong.
The journey that led to the basket may have included an on-site search result, a category banner, an email, a coupon and a video impression off-site. Treating the final nudge as the whole story misprices everything that came before it. It’s like giving all the credit for a goal (largely made possible by teammates) to the footballer who scores, while ignoring the players who worked to provide a simple tap-in.
Retail media has expanded the canvas for brand exposure and measurement. Customer journeys now span a retailer’s own websites, plus off-site environments such as Meta, DV360 (Google’s demand-side platform) and CTV (Connected TV), alongside till-level coupons, email and in-store media. A last-click lens cannot see that system, so it cannot value it.
Only accurate multi-touch attribution (MTA) can reveal the full story.
Closing the loop without closing our eyes
Closed-loop measurement matters because first-party insights let us connect exposure to outcome with confidence. Where a unique identifier such as a loyalty card number is present at purchase, on-site and in-store exposures can be deterministically tied to transactions, thus creating a ground truth for what actually sold.
Off-site is slightly different. Partners cannot always share person-level signals, so exposure is often visible as aggregate reach and frequency for a defined audience. But even where person-level signals are unavailable, there are still ways of modelling the outcomes accurately to complement the view of attribution as a whole.
These include audience-level aggregated exposure rates, historical performance and campaign context to infer off-site contribution in a balanced way. So in practice, the MTA model is overwhelmingly deterministic, with a measured probabilistic layer where required.
From channels to contributions
The point of MTA is not to crown a new “winner,” but to move from channel reporting to contribution analysis. The model should evaluate how each touchpoint shifts purchase probability, so an end-cap can keep its rightful influence, while upper-funnel exposures (such as a paid social impression two weeks prior) also receive proportionate credit. Essentially, what’s needed is transparent and real-time reporting. This broader view stops under-investing in the media that quietly set up the score.
Just as important is comparability. Brands have long asked for a consistent way to measure on-site search, email, coupons, in-store screens and off-site reach within one campaign and across successive flights.
A unified model lets marketers see relative performance and the synergistic effects of combining channels, rather than treating them as isolated lines on a spreadsheet. In many categories, combinations might well outperform solos, so a model that can quantify the lift from “togetherness” would help brands understand precisely how this occurs.
Time matters: getting windows and overlaps right
We all know that measurement windows should not be one-size-fits-all. A family pack of cereal, a premium skincare refill and a small kitchen appliance each live on very different purchase cycles. Good MTA flexes windows by category and objective, tests those assumptions against observed behaviour and refuses to measure in-line with whatever a third party’s default happens to be.
Overlaps matter too. Real campaigns stack: brand activity runs alongside promotions, seasonal pushes meet always-on search and multiple vendors may be live at once. An MTA system should split credit across concurrent campaigns and channels in line with their marginal effects. Brands should know how overlapping campaigns are handled and whether they can see cross-campaign credit allocation.
Brands should also be able to quantify whether channel combinations outperform single-channel runs and by how much.
The model customer
The industry’s long attachment to last touch is understandable. It feels simple and it flatters the most visible media. But if we want to finance what truly works for customers and fund the right mix of on-site, off-site and in-store experiences, we need to credit the whole play, not just the tap-in at the end.
Better models are not about chasing prettier charts, they’re about relevance and optimised decisions. If we can see which sequences of touchpoints truly raise conversion for a given customer segment, we can dial down wasteful impressions and dial up the messages that help. Then shoppers can experience fewer blunt prompts and more timely, useful nudges.
True multi-touch attribution works by pinning deterministic purchase truth to a cross-channel exposure map, then weighting touchpoints by their real contribution. Done well, it delivers accurate, timely measurement and it closes the loop without closing off important parts of the journey.
With attribution technology now being deployed across retail media networks, the conditions finally exist to value every influence in proportion to its real impact on sales: both retailers and brands can begin to align investment decisions around verified contribution rather than proximity to purchase.
