Ranking studio FAQs

This guide answers frequently asked questions about the Ranking studio.

Ranking studio application

How does Ranking studio differ from Algo Weight Customization?

Ranking studio lets you define each signal's weight relative to the others. Algo Weight Customization only lets you add a boost on top of the signal weight the algorithm has learned.

Ranking studio lets you create your own algorithm by customizing one or all of the signals that power search and category ranking. You can also add signals to the mix to build your own algorithm.

Are there changes to the Search and Category API response format?

No. The Search and Category API response format is unchanged.

Does the feature require changes in the feed or pixel?

No additional integration effort is required on your end.

Customers with Bloomreach advanced base ranker (ABR)

If I change the search ranking behavior, how does it affect other features?

This table summarizes the impact on other features with ABR:

Feature	Expected impact
1:1 personalization	No impact.
Content search	No impact.
Dashboard synonyms	No impact.
Existing overrides with legacy ranking	Continue to apply.
Facet ranking	No impact.
Keyword search precision modes	No impact.
Lookups	No impact.
Merchandising operations (boost, bury, slots, add to recall, and so on)	Merchandising rules continue to apply on top of ABR.
Product grid insights	Shows the active ranking algorithm per query, next to the query context. Product grid insights shows the scores behind each product's position with Bloomreach advanced base ranker (ABR): Performance score: How the product performs on behavioral signals, with relevance already factored in. Relevance score: How well the product's content matches the search term. Total score: The product's final ranking score that determines its position in the results. Combines the performance score (which already includes relevance) with the merchandising score.
Query relaxation	No impact.
Recommendations and Pathways	Ranking changes reflect in the results.
Redirects	No impact.
Relevance by segment, Real-time segments	ABR supports Real-time segments and Relevance by segment. Existing customers: If you already use segmented signals, those segmented experiences continue when ABR is enabled, so segmented signals apply from day 1. New customers or customers enabling segmented signals for the first time: Segmented signals become available after an initial data collection period of 30 days.
SEO	No impact.
SKU Select	No impact.
Spell correction	No impact.
Variant slicing	Not supported.

How is ABR different from Algo Weight Customization?

Algo Weight Customization adds a boost on top of the signal weights the algorithm has already learned. ABR, available through Ranking studio, lets you define each signal's relative weight from the ground up and add custom catalog-specific signals.

How does signal weight customization work with ABR?

Algo Weight Customization continues to work with ABR. Boost settings and any weight customizations you've already set still apply on top of ABR.

To stop a performance signal from influencing ranking, set its weight to 0 in the Ranking studio. ABR doesn't support turning off performance signals. If you don't have access to the Ranking studio, contact your Customer Success Manager to set the weights for you.

Can I A/B test Bloomreach advanced base ranker?

Yes. You can run an A/B test from the application or by using the frontend API parameter.

Does Bloomreach advanced base ranker require feed or pixel changes?

Ensure that you're sending pixel data and custom signals in the required format. These are crucial for successful algorithm training.

What is the custom algorithm limit?

You can create up to 5 custom algorithms per catalog.

Is the ABR algorithm specific to my catalog?

Yes, given enough data is available. For the Bloomreach advanced base ranker (ABR), training uses your catalog's own shopper behavior to determine signal weights unique to your store. It is personalized and calibrated to your business by default.

If I manually change the signal weights, are those static?

Yes. If you manually override signal weights, you are no longer using the machine-learned algorithm. It becomes a custom algorithm that you can edit via the application.

How current is the ABR algorithm?

ABR is trained on your most recent 30-40 days of behavioral data. Bloomreach recommends retraining after significant changes to the catalog, custom signals, or shopper behavior.

Does ABR retrain continuously?

No, ABR doesn't auto-retrain. It is trained when triggered from the application and stays live until the next retrain. However, when you do retrain, it uses two overlapping 30-day windows (the last 10-40 days) to maximize stability and protect against anomalies in weekly data.

If you want to retrain, it is recommended to wait 30–40 days after any major event, such as a sale, promotion, or catalog change, to remove any bias from the training data.

Customers with legacy ranking algorithms (Classic ranking and Bloomreach-Optimized search)

If I change the search ranking behavior, how does it affect other features?

This table summarizes the impact on other features:

Feature	Expected impact
Content search	No impact.
Dashboard synonyms	No impact.
Facet ranking	No impact.
Keyword search precision modes	No impact.
Lookups	No impact.
Merchandising operations	Merchandising operations like boost, bury, slots, and add to recall continue to apply. Review the impact of existing rules if there's a significant deviation in the curated algorithm's signal weights.
Product grid insights	You see the new experience, with a UI indicator that the new algorithm is in use.
Query relaxation	No impact.
Real-time segments	Ranking is optimized based on the newly added signals and the existing segment signals.
Recommendations and Pathways	Ranking changes reflect in the Recommendations and Pathways results.
Redirects	No impact. Redirects continue to function the same.
Relevance by segment	Ranking is optimized based on the newly added signals and the existing segment signals.
SEO	No impact.
SKU Select	No impact.
Spell correction	No impact.

Can I A/B test the ranking experience?

Yes, you can A/B test the algorithm using the application and the frontend API parameter.

Is the Bloomreach-Optimized search algorithm specific to my catalog?

No. The Bloomreach-Optimized search algorithm is trained on aggregated data from multiple customer catalogs across Bloomreach. It isn't personalized to your catalog by default.

If you want a ranking model tailored to your catalog, you have two options:

Use the Bloomreach advanced base ranker (ABR), which trains directly on your catalog data and is the recommended base ranking algorithm.
Add a custom signal to trigger catalog-level training in Ranking studio, which trains on your catalog's behavioral data.