Glossary
Learn the definition of commonly used terms in Constructor documentation
A/B testing
A/B testing is a practical application of the scientific method where you compare two versions (A and B) of something, like a webpage or feature, to see which performs better. You start with a hypothesis about which version will work best, run experiments showing each version to different users, collect data, and then analyze the results to confirm or revise your hypothesis based on which version leads to better outcomes.
Constructor supports A/B testing to help retailers optimize manual settings. Merchandisers can conduct A/B tests for global rule configurations (Browse and Search) as well as different Recommendations strategies.
Allowlist
A list of information (for example: products, data points, IP addresses, etc.) that are permitted in a certain end destination (for example: search results, API access, etc.). Often this information meets a rule or filter requirement or is used to gain more control and customization when meeting a specific business objective.
Allowlists exist throughout Constructor’s products. For example, retailers can apply allowlist rules globally across multiple user actions. A retailer could create an allowlist rule that ensures only in-stock items appear across all search and browse requests. Allowlist can also be applied to behavioral tracking, ensuring that only relevant data is processed and used for personalization and recommendations.
API
An API (Application Programming Interface) is a means for different software applications to communicate and interact.
Constructor provides multiple REST APIs that allow integrating parties to return results from our diverse product offerings.
For example, using our API, parties can:
- Add, update, or remove items, variations, and/or item groups in a product catalog
- Configure product facets and metadata returned in results
- Create collections of products based on dynamic rules
- Define search synonyms, sort options, and searchabilities
- Customize results through searchandising rules (boosting, burying, and slotting products)
API key
A public identifier used to specify an index in API calls.
With an API key, Constructor makes it so retailers’ shoppers get results from Search, Browse, Recommendations, Autocomplete, or the AI Shopping Assistant endpoints when they interact with those experiences on a website.
For example, when a user enters a search query, the API will retrieve results from the retailer’s Constructor index and return those results on the search results page.
Attributes
Also known as: fields, metadata, properties
A parameter within a product feed that defines a product or its variants (for example: color, brand, fit, size, etc.).
Attribute based slotting (ABS)
The dynamic placement of items in a search or browse experience based on their attributes. Rather than manually selecting individual items to slot, retailers can define rules based on item attributes (such as color, brand, size, etc.), and Constructor will automatically select which items to display in the slots based on those attributes.
Beacon
A snippet of JavaScript configured on a retailer’s website. This beacon captures anonymous information about users' clickstream activity, and this data is sent to Constructor’s servers where it is used to train our AI models to help users find the products and content they are looking for on a retailer’s site.
Boost
Apply a boost rule to an item or attribute to increase the ranking of affected products in the results. A boost can be weak, moderate, or strong, and is intended to augment the rankings generated by Constructor, not override them. This can be done globally, by category, collection, or by search term.
For example, a merchant can boost a specific category of product, like sweaters, to help them gain more visibility in search and browse results.
Blocklisting
Remove a product from a results set either globally or by a specific search query, category, or collection. Blocklisting will remove a product from your site but will still retain it in your catalog.
For example, you can blocklist an overselling product to minimize impact quickly.
Browse
The ability to navigate through a product catalog by exploring categories without entering specific search queries.
Browse page
A non-search product listing page (for example: a category page listing mascara products or a page listing all blue rugs).
Browse pages can be powered by categories (groups) for structured and consistent hierarchies, or represent dedicated landing pages derived from product facets (color or brand).
Bury
Apply a bury rule to an item or attribute to decrease the ranking of affected products in the results. A bury can be weak, moderate, or strong, and is intended to augment the rankings generated by Constructor, not override them. This can be done globally, by category, collection, or by search term.
For example, you can bury products that are out of stock or you can boost products that are on sale.
Campaigns
A means of applying a set of predefined rules across multiple search queries, browse categories, or facets pages to influence what results or recommendations are returned.
For example, a retailer could run a Halloween campaign with a boost searchandising rule to a variety of Halloween-related search queries to push costumes, candy, and Halloween decorations to the top of the search results page
Catalog
A digital file, like a .CSV, that contains product or content information. This can include product name, ID, description, price, image, variants, and more.
For a language-specific search experience, each language will require its own catalog.
Category
Also known as: group, Browse
A classification or grouping of items within a catalog that share common characteristics. Categories organize items in a way that makes it easier for users to find what they are looking for.
Categories can be hierarchical, with subcategories nested under parent categories. For example, a fashion category might have several subcategories, such as sleepwear, denim, and outerwear.
Clickstream
The trail of activity a shopper leaves as they interact with your website, including things like clicks, add-to-carts, and purchases, but also negative signals like items that were scrolled past. Each interaction—from the time they arrive at your site to the time they leave—is recorded and stored in the clickstream anonymously.
Clickstream data can be used to discover a wide variety of information, such as which page shoppers visit, what products shoppers convert on, and what products shoppers add to or remove from their cart.
Cognitive embeddings (CES)
CES identifies relevant and attractive items by understanding the underlying intent and meaning of a user’s search query. It’s an improvement upon traditional keyword matching as it does not rely on exact matching to identify results. Instead, it returns relevant, neighboring results. Users receive these suggestions rather than a “no search results found” experience.
For example, a user searches for “bedside table.” Traditionally, a bedside table is a smaller table that is lower to the ground to be level or near level with a bed. With CES, results might include smaller tables (like children's tables) as well as “side tables.” It’s understood what the user is trying to find, and the results show items close to a bedside table to deliver products the user will find relevant.
Connector
A connector is a tool that fetches, transforms, and syncs data from a third party into a format that Constructor can understand. You can use the Constructor Connect CLI—a command line interface—to interact with the connectors and manage your catalog integrations.
Constructor offers several connectors to enable different use cases, from syncing product feeds and content data to facilitating front-end integration. There are partner-specific connectors, such as Shopify and Akeneo, and a general-purpose Omni Connector.
Content Search
The ability for a user to search for any type of information or content that is not a product (for example: blog articles, buying guides, recipes, etc.).
Content in search
The ability for a retailer to include content pieces (for example: blog posts, recipes, etc.) as part of the results for a search query or browse category.
For example, if a user searches for “camping tips” on an outdoor retailer's website, they would see results that include products like tents and sleeping bags but also blog posts or guides on how to choose the right camping gear or popular camping locations.
Displayability
Displayability controls whether a field, such as a facet or metadata, will be returned in an API response for Search, Browse, Collections, and Recommendations requests.
Retailers can require an authenticated token to display those fields they wish to remain inaccessible via API, such as profit margin. Also, they can make it hidden and only request the information when needed.
Exact searchable
Exact searchability requires a user to enter an exact match for the associated item to be returned.
Event
An action a shopper takes on your app or website. Common events include, but are not limited to:
- Product page view
- Category listing page view
- Search results page view
- Conversion page view
- Add to cart event
- Search and suggest events
Facet
Also known as: filter, dimension, attribute
Product attributes belonging to a set of products on a search or browse page. Facets allow a shopper to further refine results to a specific category or attribute.
Constructor will return facets based on how relevant they are to the query. For example, if the search is toys for a 5-year-old, Constructor will not return an age facet since the query already specifies an age.
Fuzzy searchable
Queries will allow for spelling correction, stemming, etc. when returning results
Fuzzy searchability allows for more flexibility for spelling corrections and stemming (for example: gardens, gardening, garden) when returning results.
For example, a user wants to purchase a “hoodies” but searches for “hoody” or “hoodied” or “hoodies.” With fuzzy searchability, the product would be included in the result set given the interpretation of the search term.
Hidden facet
A facet hidden from use as a filter but still available for searchandising. It also appears in the API response when the specific API parameter is requested.
Index
Also known as: environment
An index is a container for the search experience provided to a company. These containers store configuration information for item groups, synonyms, catalogs, etc. Each index has many sections, and each section has a single catalog. One company can have many indexes.
Item
A product list in a retailer’s catalog file.
Natural Language Processing (NLP)
Natural Language Processing (NLP) allows customers to search using natural language queries, such as “snacks kids will like” or "red dresses under $50," which can improve the accuracy of search results.
Constructor uses a variety of NLP techniques to improve search accuracy. Additionally, our query expansion includes techniques like spell corrections, auto-generated synonyms, acronyms, and more to construct a firm understanding of buyers’ intentions so that the most attractive results are returned.
One-way synonym
A synonym that expands a result set to include items that fall under the category of a search term.
For example, one-way synonyms expand recall for parent terms only (e.g, Fruit -> banana, grapes). Users searching “fruit” will see “fruit”, “bananas”, and “grapes” products, but users searching “bananas” won't see “grapes” or “fruit” products unless some other synonym has this effect.
Pod
Also known as: recommendations pod
Pods represent a specific location on a page or page template where recommendations can be rendered. A pod must have at least one strategy present in it. A page may have one or more pods. Each pod will be used to serve a unique goal by implementing one of the Constructor strategies. Strategies represent a particular algorithm used to generate recommendations and are ready to use right out of the box.
With Constructor, a pod id is necessary to retrieve recommendation results as it allows our APIs to know which pod to retrieve recommendation results for.
Protected facet
A facet hidden from use as a filter and from the API response. Protected facets will only appear in the response if an authentication token is passed through parameters.
This is useful when the data is sensitive but a retailer still wants to be able to use searchandising tools with it. For example, product margin data. A retailer may want to boost products with high margins but they wouldn’t want that information visible on their site or in the API response.
Redirects
Redirects customers to the most appropriate experience, which is sometimes outside of the search or browse journey altogether. Multiple matching criteria can be defined and used to trigger a redirect to any URL of the customer’s choice.
For example, customers looking for returns or store hours in your search bar are not searching for products. With a redirect, you can ensure users land on the page with your return policy information or store location(s) and hours.
Searchabilites
A group of settings in Constructor that dictate how metadata being passed from a catalog is used in search queries.
Searchandising
Also known as: search merchandising
The ability to override Constructor's algorithm and manually rank products according to business objectives. Searchandising rules can be turned on and off, or schedule to apply for a specific time period.
With Constructor, searchandising also refers to merchandising for browse pages, collections, quizzes, campaigns, and recommendations.
Segment
A user segment, or a group that a user belongs to. Often used to distinguish between different groups of users when applying searchandising rules.
Session
Any continuous period of activity or interaction with a site or app. Constructor defines continuous as no more than 30 minutes between actions.
For example:
A shopper opens an app, adds something to their cart, but gets distracted for 15 minutes before completing their purchase.
This is considered one session since both actions (adding to their cart and purchasing) happened within 30 minutes.
If the shopper had been distracted for 45 minutes, then it would be two sessions since the actions happened in two separate 30-minute sessions.
Slotting
Also known as: pinning
Placing a product in a predetermined position in search and browse results. This allows a retailer to ensure the product shows in the first slot (or first row) of a related results page. Slots are often used for promotional purposes, to highlight certain products, or to implement merchandising strategies
Success event
A custom event that indicates a user has successfully completed a particular action. Example success events include add to cart, subscribe, add to wishlist, favorite, like, and more.
Retailers can share custom success events with Constructor to improve ranking and personalization. We work with retailers to configure these success events on a case-by-case basis. These events help our algorithms gain more insight into important actions users take on a retailer’s website and/or mobile app, which in turn helps us better optimize for a retailer’s key metrics and business goals.
Synonyms
A term that means the same or about the same as another term. Synonyms offer retailers flexibility in addressing gaps within a search experience.
For example, retailers can identify similar terms (couch and sofa) or regional terms (wellies for rain boots) to improve a user’s ability to locate a product.
Two-way synonym
A synonym that can be used interchangeably to account for terms that mean the same thing.
For example, "oj" and "orange juice." Users searching for "oj" or "orange juice" will see both "oj" and "orange juice" products.
Variation
An option for a product, such as size or color (or even a different price that’s only available to some users). Constructor supports the concept of products with variations.
For example, if you're selling a sweater that comes in 5 sizes and 10 colors, the sweater itself would be considered the product and each size/color combination could be a separate variation.
Updated 11 days ago