Dynamic re-ranking
Automatic, real-time reordering of search results based on behavioral signals, context, and business rules, so each user sees the products most likely to convert at the top.
Definition
Definition
Every search engine does two things: it finds products that match the query (matching), then decides in what order to show them (ranking). In a conventional internal search engine, that second step is calculated once and applied equally to everyone. The list is static: whether you search or someone else does, the order is the same, based on text relevance and maybe a global popularity score.
Dynamic re-ranking breaks that rigidity. Instead of a single criterion, it combines multiple signals recalculated for each search, in real time:
- Behavioral signals: what the user has viewed, added to cart, purchased before, and which price ranges they tend to browse.
- Contextual signals: device (mobile vs. desktop), geolocation, time of day, seasonality.
- Catalog signals: stock availability, product margin, recency, historical conversion rate per item.
- Business rules: active promotions, inventory clearance needs, prioritization of own-brand or sponsored products.
These signals feed a scoring model that assigns a rank to each product for that specific query, at that specific moment, for that specific user. Products are reordered by that score before being displayed.
What separates a well-built re-ranking system from a basic one is the feedback loop: every click, purchase, and abandonment updates the signals, and the model adjusts positions in future searches. It isn't a one-time configuration; the system learns continuously.
What it's used for
What it's used for
89% of online shoppers only look at the first page of results. If the product best suited to a given user appears at position 15 instead of position 3, they will likely never see it. Dynamic re-ranking addresses that directly. By placing the products with the highest probability of conversion in the top positions, calculated for each user and each query, it produces measurable effects: higher click-through rates on search results, more products added to cart, and more completed purchases. It also lets merchants align user relevance with commercial objectives: clearing slow-moving stock, launching a new collection, or surfacing higher-margin items, without degrading the search experience.
Doofinder
Doofinder's Boosting tool is a set of features that change the relevance of products returned in search results. It operates in layers: doofinder Boosting Rules: attribute-level rules that promote or penalize results based on feed values (brand, category, availability, price range). Each rule can be tied to specific search terms (exact or broad match) or applied globally.
- Manual Boosting: product-by-product adjustment of the boost factor, directly from the admin panel. A factor above 1 moves a product up; below 1 moves it down.
- Automatic Boosting: Doofinder analyzes the most clicked and purchased products over a given period and boosts them automatically, on the assumption that frequent user selection signals relevance.
- Feed-level boosting: the boost factor can be included directly in the product catalog feed, allowing prioritization to be automated from the product management system itself.
These layers work alongside Doofinder's AI personalization, which adapts results based on each visitor's browsing and purchase history.
References