Skip to main content

Boost and Bury

Boost and Bury rules raise or lower products in search results and collections by adjusting their relevance scores. You can apply them to a specific query or collection, or set global rules that apply to all searches or all collections.

Overview

A boost/bury rule is a score modifier: it names a numeric field on your products and a weight. The weight is applied to that field's value and folded into the product's relevance score, so products with a higher value in that field move up (a positive weight) or down (a negative weight). Nothing is removed from the results, and no product is forced into a fixed slot.

This is the key thing to understand before you configure anything: a rule targets a field, not a set of products matching a condition. To boost high-margin products you boost the margin field, and every product is affected in proportion to its own margin. There is no "if margin is greater than 30%" condition, and no percentage impact.

To act on a subset of products instead, use a different control:

You want to…Use
Rank products by how much of some attribute they haveA boost/bury score modifier
Restrict a trigger's results to certain productsA filter rule
Remove specific products entirelyExclusions
Force specific products into fixed positionsPinning
Favour recently added productsA recency rule, see New arrivals

Use Cases

  • High-margin products: boost the margin field so higher-margin products rank above comparable ones
  • Inventory management: boost the stock-level field so well-stocked products rank first, and bury or filter out what you cannot ship
  • Popularity: boost a click, view or sales-count field so proven products surface earlier
  • Ratings: boost an average-rating or review-count field
  • Category balance: boost a field that scores how strongly a product belongs to a category you want prominent

Any field you boost must be numeric and present on your products. A rule pointing at a missing or non-numeric field has no effect.

How to Configure Boost and Bury

Step 1: Navigate to Merchandising

  1. Go to the Merchandising section in your Marqo dashboard
  2. Select your index from the dropdown

Step 2: Choose Your Scope

You have three options for where to apply boost/bury rules:

Option A: Specific Search Query

  1. Select Search Merchandising
  2. Enter a specific search query (e.g., "running shoes")
  3. Configure boost/bury rules that apply only to this query

If the query belongs to a synonym group, author the rule against the group's parent phrase. See Synonyms and Taxonomy Tools.

Option B: Specific Collection

  1. Select Collection Merchandising
  2. Select a specific collection from the dropdown
  3. Configure boost/bury rules that apply only to this collection

A collection rule applies when a shopper is browsing the collection. A search made inside a collection resolves against your search rules instead.

Option C: Global Rules

  1. Select Global Rules
  2. Choose whether to apply to All Search or All Collections
  3. Configure boost/bury rules that apply globally

Step 3: Add Score Modifiers

For each rule, choose the field and the weight:

  1. Select the field to rank by. It must be a numeric field on your products.
  2. Set the weight. A positive weight boosts, a negative weight buries. The magnitude controls how strongly the field influences the final score relative to relevance.
  3. Add more fields if you want to rank on several signals at once. Every modifier in the rule applies; there is no priority order between them, and none of them overrides another.
  4. Save and publish the rule.

Because the weight is applied to the field's raw value, the right magnitude depends on that field's scale. A field measured in thousands needs a far smaller weight than a field measured from 0 to 5 for the same effect. Start small, publish, and compare results.

How Multiple Rules Combine

Rules do not compete for priority. They combine:

  • A trigger's own modifiers and the global modifiers both apply.
  • Where the same field is modified by both, the two weights are added together. A global boost of 0.5 on margin plus a query-level boost of 0.3 on margin behaves as a single boost of 0.8.
  • If you do not want a trigger to inherit the global modifiers, turn off global score modifier inheritance for that trigger. The trigger's own modifiers then apply on their own.

The same inheritance switch exists separately for filters.

Filters Versus Boosts

A filter rule attaches a condition to the trigger and removes everything that does not satisfy it. This is where attribute conditions live, and it is the right tool when a product should not appear at all rather than appear lower.

Use a bury when the product should still be reachable, and a filter or exclusion when it should not.

New Arrivals

Ranking by recency is a separate control rather than a boost, because it needs a date field and a decay curve rather than a single weight. A recency rule names the date field, the window it applies over, and how sharply relevance falls away outside that window. Use it instead of trying to express "added in the last 30 days" as a boost condition.

Behaviour at Search Time

  • Score modifiers change ordering only. Nothing is added to or removed from the result set.
  • A request that applies its own sortBy suspends score modifiers, along with the other rules that reorder results. An explicitly sorted request is taken as the stronger ordering instruction.
  • Rules can be scheduled. See Scheduled Rules to make a boost active only between a start and an optional end time.
  • Published changes take a few minutes to reach live traffic.

Best Practices

  1. Start small: pick a low weight, publish, and measure before increasing it
  2. Know your field's scale: the same weight behaves very differently on a 0-to-5 rating and a five-figure inventory count
  3. Prefer one strong signal: several competing modifiers on one trigger are hard to reason about
  4. Monitor performance: track conversion rate and revenue, not just position changes
  5. Use global rules sparingly: they apply to every search or collection and stack with everything else
  6. Combine deliberately: use pinning for exact positions and exclusions for removal, rather than an extreme bury