Attribution Overview

At Topsort, we specialize in doing transparent and targeted advertising that not only makes advertisers feel empowered by their ad campaigns but also safe that the reported metrics are reliable and trustworthy. A cornerstone of this reliability comes from Topsort’s advanced Ad-Purchase Attribution system, which tracks purchases and matches them to customer interactions with Topsort-powered ads. This fundamental piece of the Topsort ads ecosystem provides a feedback loop for both the advertisers and Topsort’s Bidless algorithm, which is in charge of optimizing ad campaigns on behalf of the advertisers.

Through Topsort’s APIs, advertisers can track and match attributing and attributable events, ensuring a complete view of the campaign’s impact and effect on conversions. We also offer flexible attribution logic so that marketplaces can choose the best-suited method for their business case. A key aspect of this flexibility is that marketplaces can define their attribution matching model, deciding between multiple attributing events directly linked to a user and a specific campaign, such as clicks or impressions, and defining the attribution window — the time frame in which user actions, like purchases, can be attributed to a campaign. By offering this level of customization, Topsort empowers marketplaces to tailor attribution models to fit their unique needs and performance goals, whether they want to optimize for awareness, engagement, or direct conversions.

Attribution types

Click attribution

Click-based attribution is one of the most commonly used mechanisms in retail media. The main idea is that a user’s purchase is attributable to a specific ad campaign if the purchase happened within a predefined window after the user clicked on the ad. In this case, the attributing event is a user clicking on an ad associated with a product, and the attributable event is the user’s purchase of that product. The predefined window is generally no longer than 14 days.

Impression attribution

Impression attribution relaxes the click condition and only requires the customer to have seen an ad before buying a product to attribute the purchase to the ad campaign. We use the user viewing an ad associated with a product as the attributing event. This model is particularly suitable for banners, often displayed on the homepage of websites and apps.

Halo attribution

The two models described above only consider same-product interactions between users and ads. While this might be the most obvious way of attributing purchases to ad campaigns, it does not always capture the full value delivered to the advertiser. A significant amount of advertiser revenue will come from products related to the main advertised product.

For example, suppose a brand creates an ad campaign to promote a new product. It is likely that this investment not only attracts people’s attention to the new product but also reminds them of old products that they know and love. Another example is marketplaces where sellers have their shops. An ad for a specific product is often a gateway for customers to visit and explore the seller’s catalog. Both cases will positively impact revenue beyond the advertised product, which is not detectable when attribution is constrained to same-product clicks and impressions.

We developed the Halo Attribution model to mitigate these limitations by replacing the same-product constraint with a same-vendor constraint. This model can report to an advertiser exactly which products experience additional revenue from ads, independent of whether the product was linked or not to an ad campaign. The Halo Attribution model can be set-up considering clicks or impressions as the attributing event.

Configuration

The default configuration is Click attribution with a 14-day attribution window. Both the attribution model and window can be modified to match the marketplace needs.