Search ranking with ABR
This guide explains how Bloomreach advanced base ranker (ABR) ranks products for keyword search. It covers the signals the algorithm uses, how they're grouped, and how ranking scores combine to order your results.
NoteThis guide covers keyword search with ABR. If you are using legacy ranking, read the guide on legacy ranking.
How ranking works
When a shopper runs a search, Bloomreach ABR algorithm returns the matching products and ranks them to help optimize for your business KPIs.
ABR scores each product using key signals, then orders products from the highest score to the lowest. ABR learns how much each signal matters from your catalog's own behavioral data, so the ranking reflects how your shoppers browse and buy.
ABR organizes ranking signals into 3 groups:
- Query search performance: how a product performs for the exact search term.
- Overall product performance: how a product performs across your whole store, adjusted for relevance.
- Relevance signal: how well a product's content matches the search term.
The algorithm weighs these groups together for each product.
Query search performance
These signals measure how well a product performs for a specific search term. ABR uses proportions of activity (shares) rather than raw totals, which keeps rankings accurate whether a product gets 10 or 10,000 searches.
| Signal | What it measures |
| Query product add-to-cart (ATC) rate | The add-to-cart rate for the product among shoppers who searched for a term. |
| Query product add-to-cart (ATC) share | The share of all add-to-cart events for a search term attributed to the product. Shows how compelling it is at the consideration stage. |
| Query product average order value (AOV) | The average order value when the product is purchased through a search. |
| Query product conversion rate (CR) | The conversion rate for the product among shoppers who searched for a term. This is a rate, not a share of total purchases. |
| Query product conversion share | The share of all purchases for a search term that the product drives. A high share means shoppers who search this term often end up buying this product. |
| Query product revenue per visit (RPV) | Revenue per visit for the product when shoppers search for a term. |
| Query product revenue share | The share of all revenue for a search term that comes from the product. Captures both how often it sells and the order value it contributes. |
| Query product views share | The share of all product views for a search term that the product receives. Reflects how often shoppers engage with it in results for this term. |
Overall product performance
These signals measure a product's popularity across your entire store, adjusted for how relevant it is to the current search.
ABR multiplies each sitewide behavioral signal by the product's relevance score for the query. This relevance-aware approach favors products that are both popular and a genuine match for the search.
The following are weighted by their relevance to a given keyword:
| Signal | What it measures |
| Sitewide add-to-cart (ATC) | Total add-to-cart events for the product across your store. |
| Sitewide add-to-cart (ATC) rate | The product's overall add-to-cart rate across all searches on your store. |
| Sitewide average order value (AOV) | The product's average order value across all purchases in your store. |
| Sitewide conversion | Total purchases for the product across your store. Rewards products that sell well sitewide and match the query. |
| Sitewide conversion rate (CR) | The product's overall conversion rate across all searches and sessions on your store. |
| Sitewide revenue | Total revenue the product generates across your store. |
| Sitewide revenue per visit (RPV) | The product's overall revenue per visit across your store. |
| Sitewide views | Total views for the product across your store. |
Relevance signal
This group measures the semantic match between a product's content and the search term.
| Signal | What it measures |
| Text relevance score | A computed score for how well a product's text content matches the search term. When this score is low or absent, the algorithm leans more on sitewide popularity signals. |
Segmented signals
When you use Real-time segments (RTS), ABR adds segmented versions of the query search performance and overall product performance groups. These apply the same behavioral-share and relevance-aware logic, but measured only within the active shopper segment.
-
Segmented query search performance: conversion, revenue, views, and add-to-cart shares within the active segment.
-
Segmented overall product performance: relevance-aware sitewide signals scoped to the active segment.
View their weights in the Ranking studio if you have access.
Override ranking
ABR trains on your catalog's behavioral data and learns the balance of signal weights that fit how your shoppers behave. The default calibration is already tuned to your store. You can still override the ranking with:
Merchandising rules
Adjust the default order with boost and bury rules in the application, which continue to apply on top of ABR. Review existing rules if your signal weights change significantly.
Custom signals
Custom signals bring your own business data into the algorithm — for example, sales forecasts, margin scores, return rates, or stock levels. ABR trains these alongside its built-in signals and learns how much weight each one deserves.
For guidance on choosing and formatting custom signals, see the custom-signals best practices guide.