Introducing Online LLM Burying Models—a new AI-powered enhancement that makes your search results cleaner, sharper, and more relevant in real time.

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.

We are excited to announce the general availability of Customizable User Roles with View/Edit permissions. This powerful new capability is now live in the Constructor dashboard for all customers. 

This feature allows enterprise e-commerce companies to provide every member of their team with tailored access, ensuring they have exactly the tools they need to excel—without compromising security or control.

The Challenge: Scaling Operations Securely

As an e-commerce business scales, more teams rely on a central platform for daily tasks. The Merchandising team fine-tunes results across the product discover journey, the Marketing team builds campaign-specific collections, and the Development team manages integrations and indexes. Each team has a unique focus, and providing one-size-fits-all access can create bottlenecks and introduce the risk of accidental changes to critical settings.

The Solution: Granular Control, Tailored for Your Business

Customizable User Roles solve this challenge by providing precise control over what each user can do. This functionality allows you to:

• Safeguard Critical Settings: Protect company-level information, such as API tokens and user management, by restricting access to only designated administrators. 
• Isolate Access by Index: Assign users to specific indexes, making it perfect for managing different regions, websites, or environments (e.g., Staging vs. Production) without overlap. 
• Assign Feature-Specific Permissions: Grant Edit permissions for individual features. Empower your merchandisers to manage Searchandising and Facets while restricting access to A/B Testing or Recommendations for designated roles. 
• Promote Data-Driven Decisions with View-Only Access: Create view-only roles for executives, analysts, and other stakeholders. This allows them to access valuable performance reports and analytics without the ability to make changes. 

Putting Roles into Practice 

With this level of control, you can create highly specific roles that map directly to your operational structure: 

• The Digital Merchandiser (B2B): Gets Edit access to Searchandising, Facets , Collections, A/B Testing, and Global Facet Configuration for the B2B website index, but has View access everywhere else. 
• The Marketing Specialist: Can only Edit the Collections feature to prepare for an upcoming email campaign. 
• The Regional Manager (EU): Has View access to global indexes but can only Edit merchandising rules for the eu_site index. 
• The Developer: Is given Edit access to the staging_index to test a new integration but has only View access to the production environment.

How to Get Started

This feature is now available in your dashboard. Administrators can begin creating and assigning roles today by clicking on their name in the bottom left corner, then Team Members, then switch to the User Roles tab. For a detailed walkthrough on creating roles and assigning permissions, please visit our documentation guide.

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 support@constructor.io.

We’re excited to introduce the Continuous Optimization Report in your Constructor dashboard. This report gives you full visibility into how Constructor’s tailored Continuous Optimization Plan drives measurable revenue impact for your business.

💡 What is the benefit?

Constructor has always taken a unique approach to delivering value – not through one-size-fits-all algorithms, but by tailoring search and discovery variants to each customer’s goals and shopper behavior. Until now, these results were shared mainly in your touchpoints with our Customer Success and Data Science teams. With the Continuous Optimization Report, you can now track how experiments are generating value to your business directly in your dashboard, ensuring transparency and confidence in the ROI Constructor delivers.

📍 Where to find it?

You can access the Continuous Optimization Report in your dashboard by selecting Optimization from the side menu. There you’ll see the combined annualized revenue lift from completed experiments, a list of experiments with details like revenue lift, annualized projections, and important context about each experiment's hypothesis. Click any experiment to see its setup, metrics, and outcomes – all in one place!

If you have any questions or feedback on the updated feature, please connect with your Customer Success Manager or contact us through support@constructor.io.

We know many of you have been waiting for this one — and we’re excited to share it’s finally here!

We're introducing Merchandised Page Performance (MPP) — a brand-new report that finally gives merchandisers the data to prove when manual ranking rules are helping or hurting business metrics.

Until now, rules were a bit of a black box. You could push a product up or bury something you didn’t want to show, but measuring the impact was tricky. Were these rules actually helping your business goals? 

Besides unresolved questions, merchandisers have struggled to show stakeholders whether manual merchandising decisions — like boosting new arrivals, slotting hero products, or burying poor performers — were driving results or dragging down conversion. 

With Merchandised Page Performance, you don’t have to guess anymore. The report shows:

• All pages with active manual ranking rules in one place (search, category, collections, facet pages).
• Before-and-after performance (7-day windows) for the page overall (Page Impact) and for the items affected (Items Impact).
• Easy-to-read suggestions (our little lightbulbs 💡) telling you when a rule might be helping, hurting, or ready to retire.

Here’s how customers are already using it:

• When they launch a new set of rules, they monitor the report daily for the first week to see how things are progressing.
• As a go-to tool to clean up and remove rules that no longer serve their needs.

This means less manual effort, faster insights, and more confidence that every merchandising decision is helping you hit your targets.

You can find Merchandised Page Performance today in your dashboard under:

 Analytics → Rule Performance → Merchandised Page Performance

We’re thrilled to see how you’ll use it to fine-tune your strategy. And this is just the beginning — we’re already working on making the report more granular (down to individual rules), faster  and even more powerful.

📖 Learn more in our documentation.

We’re excited to announce a new capability that makes it much easier to manage time-sensitive changes in your catalog: Metadata Overrides.

Until now, if you wanted an attribute like price to change at a specific time (for example, to match the start of a sale at exactly 9pm ET), you had to carefully time your ingestion and indexing. That was always a hassle, especially when marketing campaigns needed precise alignment with communications.

With Metadata Overrides, you can plan ahead. You simply ingest future values in advance (e.g. price_october_campaign) and then configure them to automatically override the main field (price) at the exact date and time you choose. During that window, Constructor will serve the override value in all relevant APIs (Search, Browse, Autosuggest), and once the window ends, it will automatically fall back to the original field.

💡 How it helps

• No more late-night ingestions: Set it once, and the switch happens at the exact time you want.
• Perfect campaign alignment: Prices, badges, and other metadata can flip in sync with marketing messages.
• Reliable rules: Filtering, sorting and searchandizing rules will respect the override values, so everything behaves as if the new value was the original.
  
  

🧩 How to use it

1. Ingest your future fields in advance
   
   

   • Add additional fields in your feed that hold future values (e.g. price_october_campaign, promo_price_q4, black_friday_visibility).
   • These should mirror the structure of your main fields and be in the same metadata section.
     
     

{

  "id": "coffee_machine",

  "facets": {

    "price": 10,

    "price_october_campaign": 5

  }

}

1. Configure override rules
   
   

   • Use the Metadata Override APIs to tell Constructor when to swap fields.
   • Example: from Oct 1st 9pm UTC until Oct 7th 11:59pm UTC, use price_october_campaign instead of price.
     
     

{

  "field_name": "price",

  "field_path_in_metadata": "facets.price",

  "rules": [

    {

      "start_time": "2025-10-01T21:00:00Z",

      "end_time": "2025-10-07T23:59:59Z",

      "replaced_field_name": "price_october_campaign",

      "replaced_field_path_in_metadata": "facets.price_october_campaign"

    }

  ]

}

1. Constructor takes care of the rest

   • During the configured window, all API responses will return the override value.
   • Once the window expires, the original value is automatically restored.
     
     

🔗 API Reference

• Summary page Metadata Overrides
• GET Metadata Overrides 
• POST Metadata Override 
• PATCH Metadata Override 
• DELETE Metadata Override 

ℹ️ Limitations / Information to note

• Supports Search, Browse and Autosuggest for now, Recommendation support is coming soon.
• Up to 5 override rules per catalog key.
• Only one active override per field per time window.
• Overrideable fields need to be ingested and configured at least a week before the campaign start date.
• If an override field is missing for an item, the system will return the original field value.
• Start & end time values must be formatted as an ISO date & time string, with a timezone UTC offset at the end (yyyy-MM-dd hh:mm:ss.SSS±hh:mm). An entry with a missing timezone specification is assumed to be UTC.

✅ What to do next

• Start including override fields (e.g. price_october_campaign, promo_price_q4, etc.) in your feeds now.
• Configure rules via the new Metadata Overrides APIs.

• Once configured, you can test results using the `now` parameter. It allows you to override the desired time with a specific timestamp.

  • For example, to search for "cookies" and see the results as if the query was being executed on May first at one minute past midnight:

    • https://ac.cnstrc.com/search/cookies?key=[API key]&now=1651388460
    • The now parameter requires a Unix epoch timestamp
    • This query requires an API token in addition to an API key
    • Setting a now parameter can only be set to future times
• Rules will come into effect automatically at the exact times you need.
  
  

👉 This feature removes the headache of timing ingestions and gives you reliable control over campaign/sales related changes.

Constructor now supports Directed Acyclic Graphs (DAGs) for category hierarchies—making it easier to model real-world structures where a single category belongs to multiple sections.

Instead of duplicating categories like “Wireless Headphones” to appear in both "Electronics" and "Gifts," you can now assign multiple parent paths and control which one is shown based on where the shopper is navigating from.

Benefits include:

• Cleaner catalog setup with no duplication of categories
• Flexible hierarchy modeling that reflects real-world relationships
• Better navigation control using query-time path selection

This enhancement also introduces a new groups_path parameter that lets you specify which path to render during browse requests. This results in more precise breadcrumbs and a better contextual experience for shoppers.

🔧 Example Use Case
 A single category, like Wireless Headphones, can appear under:

• Electronics > Audio > Headphones > Wireless Headphones
• Gifts > Tech Gifts > Audio > Wireless Headphones

If no groups_path is passed during a browse request, Constructor automatically uses the first parent path ingested for that category - so you’re always covered by default.

How to Structure Parent Categories

To incorporate multiple parent categories, you’ll need to update your catalog format:

✅ Do this:

• Replace parent_id: "xyz" → with parent_ids: ["xyz"]
• For multiple parents: parent_ids: ["xyz", "abc"]

⚠️ Important:

• Only include parent_ids - do not include both parent_id and parent_ids
• Even a single parent must be wrapped in an array
• Ensure all parent categories are defined before ingestion

• You can send catalogs with multiple parents in either:

  • CSV format via item_groups.csv
  • JSON format via the /v2/item_groups endpoint

You decide which path to show based on the user’s journey - with full validation and fallback behavior built in.

DAG-based category hierarchies are already available in production to all our customers. 

Related docs: Retrieve item groups, Create or replace item groups, Update item groups, Retrieve item group, JSON / JSONL feed format - item groups (categories), CSV feed format - item groups (categories) 

We're beyond excited to unveil a powerful new capability in the Constructor dashboard: Customizable User Roles with View/Edit permissions, now available in beta!

With this release, companies can tailor access like never before. Empower your teams with precisely the tools they need, whether it's managing indexes, merchandising, or launching A/B tests. It's flexible, secure, and built to scale with your company.

Why this matters

Companies tend to have multiple teams working within the Constructor dashboard. From Merchant Teams responsible for all things merchandising, Marketing Teams that use Collections for email campaigns, and Dev Teams in charge of integrations, every group uses the dashboard differently. With Customizable User Roles, you can assign access and permissions that align with each team's responsibilities without compromising control or security.

What you can do

With this release, you can now create custom roles that:

• Control access to company-level settings, such as API tokens and user management
• Allow access to specific indexes (perfect for teams that manage different regions or websites!)
• Grant edit access to individual features, such as Searchandising, Global Facet Configuration, Recommendations, and more.

You can even create a view-only role to invite more stakeholders from different departments who might benefit from the information in the Constructor dashboard, such as reporting and analytics. 

How to get started

All roles start with view-only access permission across all indexes. From there, you can assign edit access to any index and specific features within those indexes. This can be done via the Constructor dashboard. 

For more information, please visit the documentation guide.

With custom user roles, your teams will benefit from tools and information available in the Constructor dashboard without the risk of accidental actions.

We are currently rolling out Customizable User Roles with View/Edit permissions in beta. Please, connect with your Customer Success Manager or contact us through support@constructor.io to get early access.

What’s new:

We’ve upgraded Variation Slicing to give merchandisers more precision than ever.

With the new Variation Slicing Conditions, you can choose exactly which product variations to surface — without slicing all possible variants for every item.

You can now:

• Highlight only the variations you want — e.g., new styles, sale items, or seasonal colors.
• Filter variations with powerful conditions such as Item ID, color, size, or other attributes.
• Mix and match conditions to target exactly the right subset of product variations.
• Visually merchandise your PLPs by promoting only the variations that matter for your campaign.
• Use a drag-and-drop preview to instantly tailor the shopper experience.

Why it matters:

Before, merchandisers had to make a tough choice:

• Slice all variations for an attribute (risking cluttered or irrelevant product listings), or
• Avoid variation slicing altogether — missing opportunities to spotlight specific variations.

Now, you can surgically control variation visibility to align perfectly with your strategy.

 For example:

• Launch a PLP for Valentine’s Day featuring only red dresses.
• Promote only the 12-pack size of a soft drink that’s on sale.
• Push new color drops for a top-selling shoe to the top of the page.

This means more relevant merchandising, more engaging shopping experiences, and more revenue potential — without sacrificing the integrity of your PLP layout.

Learn more in the documentation.

At Constructor, we recognise the importance of data to inform merchandising decisions and track business performance, and we continue to enhance analytics capabilities, so that find the right would become more intuitive without leaving the dashboard. That's why we are excited to introduce grouping analytics by index. 

What’s new

You can now select more than one index in the dashboard’s analytics view to see combined performance metrics. This could be especially helpful in cases like:

• Analysing a specific region, where each index could represent a specific country (e.g., exploring performance of all indexes from EMEA)
• Getting insights on specific languages for the same regional market (e.g., Germany–German and Germany–English)
• Any custom grouping based on your operational structure, warehouses or team ownership

Multi-index support is available in various sections within Analytics, namely:

• Homepage
• Search
• Browse
• Recommendations
• Opportunities
• Item Analytics

Note: exporting for multi-index analytics will be supported separately in upcoming releases.

Why it matters

Previously, you could either:

• View analytics for all indexes combined, or
• Drill into a single index at a time

With this update, we’ve made it easier for merchandisers to analyse performance across multiple indexes at once — no more exporting and stitching reports together manually. That means less extra work for teams needing to analyse subsets of indexes. Now, you can group just the indexes you care about and instantly see combined insights, making reporting faster and decision‑making more precise.

Questions or feedback? Reach out to your Customer Success Manager or contact us at support@constructor.io.

What’s new:

We’ve upgraded Searchandizing Campaigns to work across multiple indexes — so you can manage cross-country and cross-environment merchandising in one place.

Now you can:

• Create once, publish everywhere — apply a single set of searchandizing rules to multiple countries, languages, or environments at the same time.
• Save hours by skipping manual copy-paste between indexes.
• Keep campaigns consistent across global markets or between dev/prod environments.
• Easily scale your strategy for regional launches, seasonal campaigns, or cross-country promotions.
• Validate faster — see the end-to-end shopper experience across all markets in just a few clicks.

Why it matters:

Before, cross-country merchandising was a time-consuming, error-prone process:

• Merchandisers had to duplicate configurations for each index manually.
• Testing and validation meant jumping between indexes one by one.
• Rule conflicts often appeared during copying — slowing launches.
• Keeping strategies aligned globally was tricky and sometimes led to outdated or inconsistent shopper experiences.

Now, Cross-index Searchandizing Campaigns make it possible to:

• Launch a new product or marketing push across all markets instantly.
• Keep branding and merchandising strategies consistent worldwide.
• Respond to market changes or promotions at global speed.

Whether you’re boosting new arrivals in all EMEA markets, aligning a seasonal campaign across APAC and North America, or syncing dev and production environments, you can now do it all from one campaign.

Learn more in the documentation.