search mobile facets autocomplete spellcheck crawler rankings weights synonyms analytics engage api customize documentation install setup technology content domains user history info home business cart chart contact email activate analyticsalt analytics autocomplete cart contact content crawling custom documentation domains email engage faceted history info install mobile person querybuilder search setup spellcheck synonyms weights engage_search_term engage_related_content engage_next_results engage_personalized_results engage_recent_results success add arrow-down arrow-left arrow-right arrow-up caret-down caret-left caret-right caret-up check close content conversions-small conversions details edit grid help small-info error live magento minus move photo pin plus preview refresh search settings small-home stat subtract text trash unpin wordpress x alert case_deflection advanced-permissions keyword-detection predictive-ai sso

Search Customizations

Search customizations are a critical part of the App Search platform. It consists of a trifecta of tools that can be used to effectively transform, alter or fine-tune the algorithmic relevance and result output for one, many or all queries issued against an engine.

Search customizations are applied using:

This guide focuses on Dashboard interactions and core concepts.
The Search guide covers the programmatic (via the Search API) use of weights and boosts in greater length.

Result Pinning

Result pinning is a query customization that maps to a single or multiple queries issued by your search experience's end user as defined in the Result Pinning area of your App Search dashboard.

Result pinning lets you perform two actions: promote a result, or prevent a result from being displayed for a specific query. Promoted results will always be returned in the API response the order it was originally manually ranked. To promote a result, simply drag a result up in the Promoted drop area. Promoted results will be returned in the API response with a score of exactly 1.0, as relevance scoring algorithms do not apply in the case of a promoted result.

Result Pinning - Promoting a result for sedan
Promoting a result

To hide a result, select any result and click the eye icon. Results will be hidden only for this query, but will still be visible for any other query unless it also has been marked as hidden for another query.

Result Pinning changes are live and propagated once saved.
No re-indexing is required, and changes are reflected immediately.

Relevance Tuning

In contrast with result pinning, Relevance Tuning is a query customization that maps to all queries issued by your search experience's end user unless some of the weights and boosts configurations are overridden at query time.

Weights

Weights function at the field level, and represents the relative importance of a field in the algorithmic formula. The higher the weight factor, the more influence a field's value will hold in the overall relevance model for all queries issued.

For instance, make should likely takes precedence over color in our car rental database example. Relative weights value vary greatly from a use case to another, and your application and data may call for something dramatically different, but since weights are relative, you will be able to properly stack fields against each other.

Weights - Increasing the influence of the make field for the overall relevance model
Adjusting Weights

It is good practice to test out a few queries using the Preview panel on the right to validate your relevance assumptions. An adjustment that may be very positive for a given query may yield mixed results for another one. Weights is a powerful tool, therefore working in small increments is advised.

Boosts

Boosts are applied based on a particular condition for any given query. Boosts can either be value-based, function-based (for numerical types) and proximity-based (for all types but text).

Boosts are useful for consistently influencing relevance based on predefined rules. For example, we may want to surface more expensive rentals at the top of our result set for generic search queries. A functional boost is an effective way to use daily_rate as a way to influence the result scoring process and therefore push more expensive results up, while leaving it to the full-text relevance to set the base score for all documents matching the user query.

Boosts - Applying an additive functional logarithmic boost to daily_rate
Applying boosts

For more information about Relevance Tuning at query time, review the Search guide's Weights and Boosts section.

Relevance Tuning changes are live and propagated once saved.
No re-indexing is required, and changes are reflected immediately.

Synonyms

Synonyms are a simple and effective way to capture and properly match on your user's intent without having to alter or modify the data indexed. Synonyms are conceptually straightforward: by creating a synonym set, you create an equivalence relationship between the terms that it contains, each term effectively inheriting the other items' relevance properties for full-text search.

For example, you may want to create a synonym set for terms like automobile, car and vehicle to properly capture a user's intent and deliver as much value and relevance.

Synonyms - Multiple synonym sets created for rent-a-car
Creating Synonym Sets

It is worth noting that one should not consider creating synonym sets for variations on a word, as App Search already handles those permutations with a robust stemming algorithm. For instance, avoid synonyms such as change, changed and changing.

Synonyms can also be set and managed via the Synonyms API.

Synonym sets are live and propagated once saved.
No re-indexing is required, and changes are reflected immediately.