The best merchandising decisions are the ones you can defend with data. You've been slotting, boosting, and tuning pages for weeks — but when the data says the page is down, which rule caused it? Now, with this major update to Merchandised Page Performance, you can finally get an answer to that question.

• ⚡ Individual Rule Performance — open any page's details drawer to see a breakdown of every rule's contribution: Page Impact (its estimated share of the total page-level metric change) and Items Impact (the direct before/after performance of the items it affected). Two lenses, one table, zero spreadsheets;
• 💨 Redesigned header — Positive, Mixed, Negative categories now show page counts and switch in a single click. Finding what needs attention just got a lot faster;
• 👴 Smarter aged rules — a rule that's been running quietly for three months and still delivering results is not a problem. It's just old. Old rules are no longer auto-flagged as negative — use the new toggle to surface them only when you want to.

Support for additonal rule types, facet configurations, and more actionable insights are already on the way.

With this update, you can now manage sort options directly in the Customer Dashboard, without using the API:

• View already configured sort options
• Create new sort options
• Update existing sort options
• Delete sort options you no longer need

How to get started

1. In the Customer Dashboard, go to Workspace → Indexes, then click Sort Options.
2. Click Create sort option.
3. Fill in the fields:
   • Sort by — internal identifier for the sort option (used by the system to reference it)
   • Sort order — whether results should be sorted ascending or descending
   • Display name — the label shoppers see in the sort dropdown (for example, “Price: Low to High”)
   • Path in metadata — the item field to sort on (for example, price, rating, or product.rating)
   • Position — the order this sort option appears in the list of sort options shown to shoppers
   • Visibility — whether this sort option should be visible to shoppers by default
4. Click Create to save.

To manage sort options in the dashboard, users require an admin role or an explicit  Sort Options permission within a custom user role. The default sort option by relevance is configured during the index setup and can be changed only via API. Changes to sort options require an index rebuild, which usually takes a few minutes. You can learn more about sort options in our API reference.

If you have any questions or feedback on the new feature or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

The dashboard just got a major upgrade for item-level analytics. Starting today, all Constructor dashboard users can explore the full onsite shopper journey for any item — not just what happened, but exactly where it came from.

• 🔀 See traffic source breakdowns — for any item, see what share of views, add-to-carts, and purchases came from Search, Browse, Collections, Facet Pages, and Recommendations, and which channel actually converts best;
• 📉 Visualise the conversion funnel — trace the full journey from pageview → click → add-to-cart → purchase, broken down by source, to spot exactly where shoppers are dropping off.
• 📋 Drill into the Pages table — a ranked list of every specific page driving traffic to this item, with per-page purchase rate and % of total purchases, down to the exact search query or category page. From any row, jump directly to Interact or Analytics to act on what you find.

Item Analytics used to tell you what happened. This update tells you why — and where to act, contributing to our mission: help eCommerce teams spend less time asking "where did this come from?" and more time acting on what they find.

We're continuing to build on this by bringing more sources, controls and ultimately connecting it with offsite traffic to give a complete end-to-end shopper journey view.

With this update, badge assignments are now fully transparent in the Interact section of the dashboard. You can:

• See which items have badges: Instantly view all badged items on any search or browse page, along with their positions in the results
• Understand why: Hover over any badge to see a detailed explanation of why that item received it (e.g., "Bestseller: Top 15% of items by purchases over the last 30 days")
• Review your shopper experience: Quickly verify that badges are being applied as expected and provide feedback when they're not

This transparency helps you understand how Constructor is enhancing your product discovery experience and gives you confidence that badges are driving value for your shoppers.

How to Get Started

This feature is available now for all indexes with badges enabled. Open a search query or browse page in Interact, and look for the Badges section in the Ranking Factors area. Click the counter to see all badged items or scroll the items grid. Hover over a badge to find explanation why an item received it.

If you're interested in enabling badges or have questions about how badges work, learn more in our badges documentation or connect with your Customer Success Manager.

With Scheduled Collections, you can now set start and end dates for a collection so it is only available to shoppers during the time window you choose.

This means you can:

• Prepare in advance: Build and QA your collection days or weeks ahead of launch.
• Launch on time: The collection becomes available automatically at your scheduled start date.
• End cleanly: Limited-time collections stop being available automatically after the end date (API returns 404 Not Found).

How to Get Started

In the dashboard, navigate to Collections, then create a new collection or edit an existing one. Set your start and or end date to schedule when the Collection should be available.

Note: Updates are evaluated on a short interval, so a collection may become available up to 2 minutes after the scheduled start time.

If you have questions about this release or Collections in general, please connect with your Customer Success Manager or contact us through [email protected].

These models identify the true meaning behind a search query in realtime and analyze how relevant to that query is each particular item within the results. By burying or removing less relevant items, the models make search results cleaner, balancing relevancy and attractiveness and helping shoppers find what they need faster.

Until recently, these powerful AI models worked quietly in the background. Now, you have full visibility in the Interact section of the dashboard into when and how LLM models kick in to improve your search results:

• See when the model is active: LLM Filtering or Burying now appears in the Ranking Factors section of Interact.
• Understand the impact: View which specific items from your top results were removed or buried by LLM models for any search query.

How to Get Started

LLM Filtering and Burying models are available in A/B-test to all our customers. Once the feature is enabled for a particular index, simply navigate to Interact, enter a search query, and look for LLM Filtering or LLM Burying in the Ranking Factors section.

If you have questions about enabling LLM Filtering and Burying models or feedback on this feature, please connect with your Customer Success Manager or contact us through [email protected].

Merchandisers can select up to 5 facets when setting up conditions for pod rules.

📍 Where to find it?

1. Navigate to Workspace → Recommendations → Add New → Recommendation Rule.
2. From there, select the pod(s) you want to set up rules for.
3. Select multi-layered conditions. 

If you have any questions or feedback on the updated feature, please connect with your Customer Success Manager or contact us through [email protected]

Collections now include modification author tracking, so you can immediately see who last updated a collection — directly in the dashboard.

For each collection, you’ll now see:

• The name and email of the last person who modified it (when available)
• A “Using API” label if the most recent update came via the API
• A “-” if the collection was last updated before author tracking began (January 2026)

Availability

Available now for all Collections customers (included as part of Collections).

Why it matters

Collections are often managed by multiple merchandisers (and sometimes in an automated way). When something changes unexpectedly, figuring out who made the last edit — and how it happened — can turn into detective work.

With author tracking, teams can:

• Resolve questions faster by seeing the last editor at a glance
• Reduce back-and-forth internally (“Was this you or me?”)
• Spot API-driven changes immediately, so automated updates don’t look like mysterious human edits

Technical limitations

• Author tracking is available only for modifications made from January 2026 onward.
• We currently track only the most recent modifier (not the full edit history).

With the recent launch of Audience Hub, Affinity, RFM, and Location segments are now available in Searchandising — expanding our Custom & Dynamic segments and upgrading how you target shoppers in rules and campaigns.

You can now pick segments based on:

• Affinity (what brands/categories/styles a shopper tends to prefer)
• RFM (purchase Recency, Frequency, and Monetary value)
• Location (geographic shopper segments)

As you select a segment, you’ll also see whether it matches the traffic volume you intend to target, so you can make confident, data-driven choices while configuring.

Why it matters

Previously, many merchandising strategies were applied too broadly (because the necessary segment wasn’t available in Searchandising or was hard to maintain), which could dilute impact and sometimes actively hurt performance when irrelevant shoppers see promoted items.

With Affinity, RFM, and Location segments:

• Target the shoppers your strategy is actually designed for, without relying on custom segment pipelines.
• Maximise incremental impact by showing promotions only where they’re likely to help.
• Avoid “segment guesswork” by checking intended traffic volume while you configure.

Example workflows

• Launch new arrivals to high-intent fans. Promote a newly arrived Nike drop only to shoppers with a strong Nike affinity, instead of pushing it to everyone.
• Run clearance with precision. Target sale campaigns to frequent buyers, or to shoppers whose recency is dropping (and may need a nudge).
• Localise merchandising. Highlight weather-appropriate or region-specific inventory to shoppers in the relevant locations.

If you have any questions or feedback on the new feature or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

We are thrilled to announce the launch of Audience Hub, a new way within the Constructor dashboard to get a clear, centralized view of your shoppers and their characteristics at scale through segmentation.

Previously, segment data was fragmented and difficult to access, making it hard to grasp the size and opportunity each segment represented. Audience Hub solves this by bringing all your meaningful segmentation data into a single location, making it an easy starting point for understanding and further optimization.

Audience Hub Main Page

With Audience Hub, you can:

• Gain a deeper understanding of your users: See shared similarities across your users, such as affinity for specific brands (e.g., Nike) or locations (e.g., Canada). See top searches, items viewed, added to cart and purchased for each segment.
• Centralize and visualize all segments: Access a single source of truth for all user segments, including geo, device, customer provided and AI generated smart segments (such as Affinity and RFM).
• Identify high-impact opportunities: Understand the exact size and opportunity that various segments represent, making it easy to decide which ones are worth further review and action through onsite and offsite optimizations.

Segments Drawer within Audience Hub provides more segment details.

We’re excited to make this available now, and we’re just getting started with segments. Keep an eye out for more to come in this space.

If you have any questions or feedback on the new feature or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

Previously, configuring a metadata field as Hidden (so it exists in the index but isn't returned in the response) required API intervention and was invisible in the dashboard.

With this update, you can now:

• Gain Visibility: See exactly which metadata fields are currently set to Hidden directly in the dashboard.
• Self-Serve: Toggle the Hidden configuration on and off without needing to reach out to dev resources and make API calls.

How to Get Started 

This feature is available now. Navigate to Workspace > Indexes > Manage Searchability & Displayability and choose a field to set up its Hidden configuration.

Important notice: Changes to metadata field configurations take effect in search and browse requests once Constructor rebuilds indexes. Rebuild duration will depend on the size of your data set, including items, facets, rules, and other indexed data.

If you have any questions or feedback on the new feature or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

Previously, customizing how the options of bucketed range facets appeared to shoppers required API intervention. Now, you can configure Display Names for these options directly in the dashboard. This allows you to:

• Add Formatting: Easily change 100-200 to $100 - $200 or 100cm - 200cm.
• Localize Language: Translate labels for different regions, such as changing "4 & up" (rating) to "4 & πάνω."

How to Get Started 

This feature is available now. Navigate to Workspace > Indexes > Global Facet Configuration and find a bucketed range facet to start customizing display names for its options.

If you have any questions or feedback on the new feature or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

Facet campaigns are here. You can now apply one set of facet rules — slotting and/or hiding facet groups and options — to many results pages at once, instead of configuring each page separately.

A single facet campaign can target a mix of:

• Search queries (e.g. “denim”, “running shoes”, “ground coffee”)
• Browse category pages (e.g. Shoes, Coffee, Skincare)
• Facet pages (e.g. “Size 7” Shoes, “Straight Leg” Denim)
• Collections (e.g. “Viral Hits”, “Kid’s Summer Shoes”)

Think of it as Searchandising Campaigns, but for facets: one place to define the order and visibility of your filters across a whole group of related pages.

Each facet campaign currently supports up to 100 queries and 100 browse/facet/collection pages (combined).

Why it matters

 Previously, keeping filters tidy and consistent meant repeating the same facet rules on every individual page:

• Hiding niche facet groups (like “Heel height”) on categories where they barely apply
• Reordering key facets (like “Brand” or “Sale”) across long trees of categories and collections
• Repeating the same updates every time your strategy changes

With facet campaigns, you:

• Configure once, reuse everywhere. Define facet ordering and visibility in a single campaign and apply it to all related pages.
• Keep filtering experiences clean. Hide facet groups and options that only matter to a tiny subset of items so shoppers aren’t scrolling through noisy filters.
• Stay consistent across parent and child PLPs. Align the filtering experience across a parent category (e.g., “Shoes”) and its children (e.g., “Boots”, “Heels”, “Sneakers”) without micromanaging each page.
• Move faster when strategy changes. Update a single campaign instead of revisiting dozens of PLPs when you change priorities (e.g. pushing “Sale” or your own brand to the top).

Example workflows

• Tidy up irrelevant facets at scale. A retailer only wants to show Heel height on women’s boots and heels—not on sneakers, kids’ sandals, or men’s trainers. They create a facet campaign, attach Shoes, Athletic Shoes, and Kid’s Summer Shoes pages and collections, and hide “Heel height” once. All included pages inherit the same tidy filter set.
• Promote own brand across a family of pages. A beauty retailer wants their own brand to appear first in the Brand facet on search results, “Makeup” categories, and “Viral Hits” collections. They add these contexts into a single facet campaign and slot their brand option to position #1—no need to repeat the rule per page.
• Align parent and child categories. A fashion retailer wants a consistent set of filters across Denim, Denim > Straight Leg, and Denim > Wide Leg, but with fewer facets on the narrower children pages. They build a facet campaign for all denim-related pages, then hide niche facet groups from specific sub-pages while keeping core filters aligned.
• Seasonal or promo-focused filtering. During peak season, a merchandiser wants the Sale, New in, and Sustainability facets to rank higher across key search queries, categories, and promo collections. They create a campaign, slot those facets near the top across all relevant contexts, and adjust positions from a single configuration when the season changes.

Learn more

• Create a facet campaign
• Configure facet rules

What is it?

A new enhancement to the Collection Creation API allows customers to make items within a Collection discoverable for specific search phrases. This means that when a shopper searches for terms like “Sale,” “Holiday Deals,” or “Thanksgiving Sale,” the items from relevant seasonal or promotional collections (e.g., “Black Friday Collection”) can automatically appear in search results with no manual items association or merchandising rules required.

This feature gives merchants a lightweight, API-only way to directly associate search queries with entire collections of items, improving recall and discoverability while reducing manual merchandising overhead.

Note: This functionality is currently available only via the API, with dashboard and UI support coming soon.

Why this matters

Previously, there was no automatic link between seasonal or promotional collections and shopper queries:

• Items in those collections weren’t discoverable by campaign-related terms, since they often lacked explicit query associations or metadata.
• Merchants had to rely on manual rules to approximate this behavior, an inconsistent and time-consuming process, often without good item-level attribution for seasonal relevance.

With Discoverable Collections, Constructor bridges this gap, letting merchants efficiently map search phrases to collections so that the right items appear automatically when shoppers search for relevant terms.

How does it work?

A new optional parameter, discoverable, has been added to the Collection Creation API.
 When creating or updating a Collection, you can set discoverable: true and define phrases and match types similar to how redirects are configured.

Each discoverable configuration includes:

• match_type (string, default to PHRASE): The matching logic for the phrase.

  • "EXACT" – The pattern must exactly match the search query.
  • "UNORDERED" – All tokens in the pattern must match the query, but order doesn’t matter.
  • "PHRASE" – All tokens must match, but the query may include additional tokens.
• pattern (string, required): The phrase pattern that triggers item discovery.
  
  

By default, discoverable is false to preserve existing behavior.

Example Request

Below is a simple example of creating a discoverable Collection that associates its items with “sale” and “Thanksgiving sale” search queries:

curl -X POST https://ac.cnstrc.com/v1/collections \

  -H "Content-Type: application/json" \

  -H "Authorization: Bearer/Basic YOUR_API_TOKEN" \

  -d '{

    "display_name": "Black Friday Test Collection",

    "id": "bf25-cnstrc",

    "filter_expression": {

      "name": "group_id",

      "value": "all"

    },

    "discoverable": true,

    "matches": [

      {

        "pattern": "sale",

        "match_type": "PHRASE"

      },

      {

        "pattern": "Thanksgiving sale",

        "match_type": "PHRASE"

      }

    ]

  }'

In this example, items within the “Black Friday Test Collection” become discoverable whenever shoppers search for “sale” or “Thanksgiving sale.”

What problem does it solve?

The Discoverable Collections Items feature eliminates the need for complex rule management and inconsistent boosting strategies. By connecting queries → collections → items directly, merchants can:

• Automatically surface seasonal or campaign items for relevant search terms.
• Reduce manual merchandising effort while improving accuracy.
• Increase recall and relevance for high-intent seasonal queries.
  
  

Availability

The Discoverable Collections Items feature is available today via the Collection Creation API. Dashboard and UI support are coming soon to make discoverability even easier to configure.

For assistance or to learn more about setup best practices, please contact your Customer Success Manager or reach us at [email protected].

The Challenge: The "All-or-Nothing" API Key

Previously, API tokens (now referred to as Legacy tokens) were company-wide, granting full read and write access to all Constructor endpoints across all your indexes. While powerful, this created potential risks. A leaked token technically, could expose all your data, and a simple misconfiguration in a script could accidentally apply changes to the wrong environment.

The Solution: API tokens with Custom Level of Access

Now, you can build API tokens with a specific, limited scope. When creating a new token, you can define:

• Indexes: Specify which indexes the token can access. This is perfect for isolating your Production, Staging, and Dev environments or managing different websites.
• Scopes: Select which API endpoints the token can use (e.g., searchandising.refined_queries, searchabilities,  synonyms).
• Permission: Define the permission level for those scopes: Read & Write, Read, or Write.
• Expiration Date: Set an optional expiration date. We recommend renewing API tokens regularly.

By limiting access to what is really needed, you reduce security risks if a token gets compromised and rule out the possibility of applying changes to the wrong environment.

Putting API Tokens into Practice

This new control allows you to securely manage complex, multi-team workflows:

• For Integration Scripts: Create an API token that has access only to scopes within the dev_index to test things before applying changes in production.
• For Regional Teams: Give your EU engineering team an API token that can Read & Write to all scopes only on the eu_site index, preventing accidental changes to the US site.

Better Together: API Tokens + Customizable User Roles

This update works hand-in-hand with our recent Customizable User Roles feature. Only users assigned with a role that has Manage API tokens permissions can set up API tokens. Users' ability to create API tokens is limited by their role permissions. An administrator can create tokens for any scope, but a user with a specific role (e.g., "US Merchandiser") can only create API tokens for the indexes and features they are already permitted to access. 

This ensures your permission model is consistent, from the dashboard UI right down to the API.

A Note on Legacy API Tokens

Your existing tokens (now labeled Legacy tokens in the dashboard) will continue to work without interruption. However, you can no longer create new tokens of this type. We highly recommend you audit your existing integrations and begin migrating to the new, more secure API tokens as part of your team's security best practices.

How to Get Started

This feature is now live for all customers. Administrators and users with the appropriate permissions can create new tokens by navigating to Integration > API Tokens in the dashboard.

For a complete, step-by-step guide, please visit our updated documentation: Generate and manage API tokens.

If you have any questions or feedback on the new feature or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

The dashboard analytics experience it taking a big leap forward. Starting today, all Constructor dashboard users can enjoy a more powerful, searchable, and segmentable analytics interface designed to make finding insights effortless.

With this update, merchandisers can:

• 🔎 Search within analytics tables across Search, Browse, and Recommendations — no more manual scrolling or exporting to find specific data points.
• ↕️ Sort by key metrics (clicks, conversions, revenue, etc.) to instantly spot what’s performing best — or what needs attention.
• 🌎 Filter analytics to uncover how different audiences behave: by geo, device type, customer-uploaded segments, Affinity, or RFM segments.

In short: you can now slice, search, and sort performance data however you need — directly in the dashboard.

Why it matters 🎯

Our users have long asked for better ways to prove impact and make data-backed decisions faster. This release brings analytics closer to that vision:

• Understand who your shoppers are and how they behave across Search, Browse and Recommendations.
• Quickly compare audience segments (e.g. iOS vs Android shoppers, Canada vs US, newcomers vs loyal shoppers).
• Focus on the moments that drive your business without leaving the dashboard or relying on exports.

These improvements make analytics both more powerful and more human, helping merchandisers move from data collection to real, actionable insight.

What’s next 🔜 

We’re continuing to expand these capabilities across the entire dashboard — bringing segment-level analytics, powerful tables, and unified interactions to more pages and reports soon.

Our mission is simple: to help merchandisers spend less time searching for insights, and more time acting on them.

The Challenge: Scaling Role Management

Since launching Customizable User Roles, we've heard feedback from large teams managing numerous brands, regions, or environments. A common challenge was managing users who needed combined permissions.

For example, to grant a developer access to both the EMEA and NA indexes, you had to create a specific "Dev EMEA + NA" role. For complex teams, this led to a long, hard-to-maintain list of "combination" roles (e.g., "Merchandiser US+EU", "Merchandiser B2B+B2C").

The Solution: Multiple Roles for Simple, Granular Control

With this update, you can now assign the existing "Dev EMEA" role and the "Dev NA" role to the same user. This approach provides powerful benefits:

• Reduce Admin Overhead: You don't need to create and manage dozens of redundant "combination" roles. Simply assign the building blocks you've already created.
• Scale with Flexibility: Easily manage permissions as your team grows and responsibilities change. Need to give a B2C merchandiser temporary access to the B2B site? Just add the "B2B Merchandiser" role and remove it when the project is done.
• Smart, Additive Permissions: Access is combined in the most permissive way. If one role grants View access and another grants Edit access for the same feature, the user will automatically get Edit access.

How to Get Started

Users with the Manage users permission can select and assign multiple roles to a single user directly from the dashboard. Navigate to the User Management page by clicking on your name in the bottom left corner, and then, Team Members. Click on the current role assigned to a user and select multiple roles in the dropdown.

If you have any questions or feedback on the update or the Constructor dashboard in general, please connect with your Customer Success Manager or contact us through [email protected].

You can now target customer-defined user segments across most searchandising rules—not just Boost & Bury. That means you can adjust ranking, recall, content, and slotting for specific audiences (e.g., "US App Users", "Loyalty Tier Gold", "Clicked Meta Ad A") without rebuilding pages or bluntly applying rules to everyone.

You can now

• Apply segments to key rules and surfaces
  • Boost & Bury — Search, Browse, Campaigns, Recommendations, Global
  • Item Slotting (incl. CSV & faceted browse) — Search, Browse, Campaigns, Recommendations
  • Attribute-based Slotting — Search, Browse, Campaigns, Recommendations
  • Allowlist / Blocklist — Search, Browse, Campaigns, Recommendations (Blocklist also on Global)
  • Content Rules — Search, Browse, Campaigns
• Preview as a segment when creating Campaigns: Impersonate a selected segment to see exactly what that audience will experience before you ship.

Why it matters

Previously, segments were available for Boost & Bury only. Teams either ignored insights or applied broad rules to all shoppers, diluting performance. With segments available across your core rule types, you can:

• Tune the same PLP differently per audience (locale, device, lifecycle stage).
• Move faster from marketing signal to on-site personalisation without duplicating pages.
• Protect performance by narrowing rules to only the shoppers they're meant for.

Example workflows

• Campaign continuity: For users who clicked a specific ad or email, slot the exact products that match the message they saw, right on arrival.
• Locale nuance: Slot a colorway or size range for only a particular country/segment without touching other markets.
• Compliance fast-track: Blocklist prohibited items for a single region while keeping them live elsewhere.
• Channel-aware content: Show different banners/copy to app users vs web visitors on the same PLP.

Who benefits

• Merchandisers partnering with Marketing: align landing experiences to campaign promises.
• Region/brand owners: tailor assortment and messaging by locale without forking your catalog.
• Compliance & ops teams: react quickly to region-specific restrictions.

Coming soon

• AI-generated segments
• Segments in Content Rules (Global)
• Segments in Variation Slicing
• Impersonation for Search, Browse, and Global

Learn more

• Configure searchandising rules
• Create a searchandising campaign

We’ve supercharged Collections so merchandisers can build, reuse, and refine PLPs with fewer clicks.

You can now

• Export item IDs to CSV from a Collection in one click—for analysis or to seed a new Collection.
• Add, edit, and remove item IDs inline in the Collections UI—no extra CSV wrangling required.
• Copy a Collection across indexes or within the same index—reuse successful setups with minor tweaks.

Why it matters

Previously, small changes took a big effort—exporting item IDs required workarounds, editing meant creating and re-uploading CSVs, and reusing a Collection often meant rebuilding from scratch.

Now you can:

• Move faster from idea to live PLP.
• Keep Collections fresh with quick, precise Item ID updates.
• Replicate proven merchandising strategies across storefronts without busywork.

Example workflows

• Re-launch a winning PLP in a new market: Copy the Collection to another index, adjust a few IDs inline, and publish.
• Clear handoff material: Provide item ID CSVs for cross-team reviews, ensuring everyone’s working from the same list.
• Tweak on the fly: Paste a handful of new item IDs straight from a spreadsheet and remove outdated ones in seconds.

Learn more

• Create a Collection
• Manage a Collection

Our system now analyses each query as it happens, identifying and removing the least relevant results on the fly. Unlike the previous offline model, which could only filter precomputed query–item pairs, this new online model works on every search call, ensuring consistent precision across your entire catalog.

Benefits include:

• Higher precision: Irrelevant or misleading products are automatically detected and pushed down or removed.
• Smarter intent understanding: The model interprets the true meaning of each query to deliver more accurate matches.
• Improved search quality: Cleaner and more relevant results help users find what they need faster.

This update strengthens your search engine’s ability to interpret intent and deliver only the most meaningful results, helping shoppers find exactly what they’re looking for with confidence.

This enhancement is now available in AB test to all our customers.

To learn more, please contact your Customer Success Manager or support team.