Raptor Site Search is an implementation of a search framework utilizing the vast library of APIs and features from the open-source cloud services of Typesense. The implementation makes it easy for Raptor customers to create, configure, and test a search index and publish it to a Typesense cluster making it available for integration on any e-commerce site.
Besides creating the infrastructure and making Typsense features available, Raptor will continuously add features that are unique for Raptor customers and built upon data and behavioral patterns of the Raptor personalization platform. The most obvious examples are wisdom-of-the-crowd ranking where best performing and/or trending products will appear at the top of the search result list, and Merchandising where you can boost certain products to the top of the search list. In addition, we are working on implementing personalized search results based on previous user behavior.
Implementing a search UI and fine-tuning the query setup on your website on top of the Raptor framework should always happen through an implementation partner offering search expertise and experience. See how to use ready-made UI components from the Typesense community here.
The illustration below indicates how to understand the split between the product that Raptor provides on one side and the strategic counselling, parameter setup and tuning, and implementation on your site on the other.
Site Search Features
The following list of features describes the current capabilities of the Site Search solution.
|
Feature |
Description |
|
Typesense Cluster |
Raptor has created a cluster in the Typesense Cloud service where search indexes will be published. More clusters will be added when scaling is necessary. |
|
Development environments |
Setup different environments for different purposes, e.g. development, staging, and production. Each environment is equivalent to one search index. You can ingest different product feeds to each index. Keep test feeds limited to a maximum of 500 products and make sure to delete test indexes if no longer needed. |
|
Product Catalog Ingest |
Product Catalog ingest succeeds through the Data Manager in a JSON or XML format. The following connectors are available for ingesting products: Streaming endpoint (recommended), HTTP*, FTP,* SFTP*, Shopify* *Ingest frequency is maximum once per hour. No custom connectors/ integrations to third-party systems |
|
Dedicated Search Schema |
Products are ingested into a dedicated Site Search schema. Focus on fields that should be searchable (e.g. name, description, brand, category), sortable (e.g. price), or used as facets (e.g. color, gender, size). Maximum number of columns: 100 |
|
Administration Interface |
Creation, testing, and administration of search indexes, synonyms, settings etc. through Raptor Control Panel |
|
Search Index |
Create search indexes from dedicated product (or content) feeds Maximum number of indexes per Control Panel Account: 5 |
|
Multiple Languages |
Multiple languages are handled with multiple search indexes. One index per language. |
|
Type-ahead |
A search is performed for each key-stroke typed by the user. This makes an effective type-ahead functionality directly coupled to the index. Use the Typesense highlighting handles to highlight the search phrase within the search results. See more here |
|
Synonyms |
Create and maintain synonyms (both one-way and multi-way) either manually or by uploading a CSV file. Succeeds in the Raptor Control Panel. |
|
Facets |
Select which columns should be available as facets in the UI. This could be a price interval slider or a brand or category selector. Facets can be different for each product category. Beware of the maximum number of columns limit (100 columns) |
|
Sorting |
Select which columns should be available as sorting parameters if tiebreak between results. You can sort on e.g. price, brand, or size. Numeric columns are sortable by default. The character limit for string columns is 500 characters due to performance considerations. |
|
Typesense parameter settings as presets |
Typesense allows many parameter settings for fine-tuning search results e.g., typos, weights, grouping, snippet length, and pagination (see the full list here). These settings are available through the Raptor Control Panel to test how they impact search results. The parameter settings can be configured, tested, and saved as presets in the Raptor Control Panel. You can call the desired preset from Typesense to apply your settings on your search index. This way you can easily edit parameter settings, test their impact, and apply them straight away to your live search solution. |
|
Search for Articles |
You can ingest both a feed of products and a feed of content (e.g. articles) through the Data Manager and create separate indexes. This allows you to display both products and articles in the same search field. A product feed and a content feed cannot be combined into the same index. During the implementation of search UI, however, it is possible to combine results of multiple indexes. Learn how to do that here |
|
Auto-correct, Typos and Zero-results |
Typesense handles typos and zero-results very convincingly, and you should only deviate from the default setup if you have very specific cases to handle. Read more here. If you need it, Typesense offers multiple settings for handling typos and zero-results. The settings can be tested in the Raptor Control Panel and saved as presets. Synonyms are also effective for handling commonly occurring typos and zero-results that might be specific for your domain. |
|
Product information in search bar |
Displaying detailed product information like description, image(s), stock status, price, etc. is possible. Information must be available and updated through the product catalog ingest in the Data Manager. |
|
Filter by |
By applying filters to your search, you can create context-specific searches on your site, e.g. only show results from a certain category when searching on a category page. You can also leave out products that are nearly out of stock. The filter by-option can be used to filter any information from any index field. The filter by-settings can be tested in the Raptor Control Panel and saved as presets. |
|
Wisdom of the Crowd Sorting (AI) |
With the dedicated recommendation strategy “All products sorted by popularity and trends for Raptor Search” search results will be ranked according to popularity and trends. The strategy has numerous weights for configuring the optimal trending period and methods. Using AI the strategy will tap into seasonal trends, micro-trends, and general trends constantly creating highly relevant search results. |
|
Merchandising |
Merchandising gives you the tools to boost products you want appearing at the top of the search results list. This could be high-margin products or a branded campaign. A Merchandising boost is a decisive factor when ranking search results at tie break between multiple products. Merchandising requires activation of the recommendation strategy “All products sorted by popularity and trends for Raptor Search” and an active Merchandising license. |
|
Account-level Pricing and Availability (B2B) |
Often, account-based pricing or special products only available to specific customers must be included in a B2B search setup. This can be supported in different ways, e.g. prices and customer IDs are contained in the ingested product feed and available per product, and the query to the index contains the customer ID. An assessment of your data structure based on the use cases you want to solve must be performed during a round of discovery. |
|
Monitoring |
The Site Search solution and all dependent Raptor systems are under constant monitoring by automized alarms and the Raptor Site Search Team. In addition, by utilizing Typesense Cloud, the Raptor Site Search solution is benefitting from the uptime and availability measures of Typesense and the vast number of websites already using Typesense as their search engine. See and follow system status here. |
|
Maintenance |
The Site Search solution is maintained, enhanced, and expanded by Raptor. We will upgrade to newest versions of the Typesense API library and utilize new features when beneficial for our customers. |
Upcoming Site Search Features
Raptor is developing new capabilities, either as further integration to Typesense or as features unique for you as a Raptor customer. Upcoming features include:
Site Search Reporting (basic numbers)
- Top search phrases
- Top zero-result searches
- Number of searches (total and over time)
Personalized Search
Vector-based personal rankings based on previous behavior (searches, visits and purchases) on your site.
🗯️ FAQ
Q: Why don't I get the results I expect when searching?
A: There are several reasons why specific search examples do not return the results you expect, and most of them revolve around how you use the tools at hand. Here's a list of suggestions where to look for solutions:
- Make sure that the right fields are made searchable. Often, the product's name, description, and brand are sufficient, but maybe the product category or attributes like flavor, ingredients, or size make sense too. Always consider if some information are better suited as facets (filtering criteria).
- How well are your products described? Searchability starts with the data that are made available to the search engine. By describing your products well in a language used by your typical customers, you can increase the quality of your search significantly.
- Typo-tolerance is built into Typesense and comes out of the box. But you can tune the typo-tolerance settings in various ways: How should spaces be treated? How many typos will be tolerated? When should typo-corrected results be shown? Do you accept typos in synonyms? These are some of the parameters that each contain multiple settings. See the typo-parameters here.
- You can do a lot of correction by adding synonyms to products that are referred to with multiple names or nicknames. With synonyms you can catch these connections by adding all names as synonyms for each other. Even though Typesense's typo-features are really strong, some products still require synonyms to be added manually. This is often the case when dealing with specialized brand names.
Q: We have more than 100 facets we want to display in our site search. Is there anything we can do about the limit of 100 columns in the Data Manager?
A: There are two ways of optimizing the management of your search solution: One is to combine multiple indexes in the same search solution, e.g. by making an index for each of your product groups. The other is to compress your data, e.g. by combining multiple columns into one (by using an array of strings).
If your need for facets exceeds the limit extensively, the Raptor Site Search solution might not be the right choice as your next provider. Typesense does not have an upper limit concerning the number of fields, so you can still use Typesense as your search engine by creating and managing your indexes in your own Typesense cluster.