The Elements of Search

infographic

The Elements of Search

Everyone uses multiple site and internet search engines each and every day. While search feels simple, it’s actually an orchestration between many different elements — ie, services and technologies. Our newest infographic breaks it down. This periodic table of search will continue to grow in time as new elements — features, APIs, and services — are discovered.

We’d love to hear what you think! Tweet your comments to @searchiohq.

View or download full size (2.5MB)

Share this
‍

Foundation

Search begins with data, indexing, schema, and basic UI elements.

Search bar: Great search starts with search... the search bar that is. Behind every search bar is your index, which is composed of site data and customer data.
Customer data: Customer data is used to personalize search and recommendations.
Site data: Data such as queries, clicks, views, and conversions can all be used to make site search smarter.‍
Indexing: The index consists of all the pages, PDFs, and products you've made searchable.
Schema: Search engines use an actual or inferred schema to help make sense of the index.

Building blocks

The building blocks enable you to deliver search at scale on different platforms.

API: Search APIs are the myriad components that enable developers to build custom search capabilities, transform data, and deliver great results for customers.
Client libraries: Search can be optimized and delivered faster with client libraries written in Node, Ruby, Java, or other specific languages.
UI libraries: UI components, particularly ReactJS, enable engineers to build custom front-end experiences.
Integrations: Pre-built integrations with CMS and e-commerce platforms like Wordpress, Shopify, and Drupal make it easy to add search to different sites.

Processing

These elements are what make search seemingly magical with every query.

Query understanding: Machines need to decode a query to understanding intent and deliver meaningful results.
‍Data transformation: Data is often transformed and enhanced as it's indexed to make it richer and easier for customers to query.‍
Natural language processing: NLP is a branch of machine learning that uses large amounts of data to process query inputs.
Algorithm: Quite simply, the search engine follows a series of steps to retrieve and rank the search results.
Keyword search: Sometimes called full-text search, a search engine will lookup results based on keyword count.
Artificial intelligence: Today, AI-based search is augmenting, and sometimes replacing, keyword search. AI search understands concepts and works well for longer, more complex queries.
Neural hashes: An AI engine that uses neural networks to compress search vectors into hashes. It's faster and more powerful than vectors alone.
Natural language search: Using NLP, keywords, and AI, natural language search provides great results for just about any query.
Signal processing: Search engines use signals such as clicks, time on page, conversions, and more, to automatically improve result ranking.
Event capture: Search engines need be able to capture and measure events. They use this data to adjust ranking and improve results.
Machine learning pipelines: For AI to work, machine learning pipelines are there to manage the data throughput and models.

Delivery

Great search experiences "just work" with these seamless delivery elements.

Speed: Slow search can impact brand perception, customer experience, and even sales. Speed isn't just a nice-to-have, it's a must have!
Vertical scale: Search needs to be able to scale, whether it's Black Friday or any other day of the year. Vertical scale is there to add more power as the number of requests grows.
Horizontal scale: Horizontal scale adds more machines to keep up with incoming queries.
User experience: Search should be designed for user experience. It impacts conversion rate, brand perception, customer service, and more.
Front-end design: The front-end — how users interact with your search box — is just one piece of a larger customer experience.
Distributed indexes: By pushing the index closer to your customers, search results are delivered even faster.
Resilience: For many businesses, search is mission-critical, so resilience features are put in place to ensure it's always running even if there's a server failure.

Features

The elements of search and discovery would not be complete without the multitude of features that can be customized for different use cases and applications.

Filters & facets: Filters and dynamic filters, also known as facets, enable users to narrow results to pinpoint exactly what they want.
Sorting: Sorting allows customers to re-order results based on different attributes such as popularity or price.
Synonyms: People know that "cars" and "autos" are the same thing, but machines don't. Search engines can have synonym management or newer AI search already knows the meaning of synonymous terms.
Domains: You're going to index one or multiple domains or subdomains for search.
Crawler: Crawlers are used to build search indexes.
Autocomplete: Autocomplete offers dropdown suggestions — which are actually probabilistic suggestions for what people are likely to want.
Instant search: With instant search, search results change in near real-time as users type their query.
Result highlighting: Search terms can be highlighted in results to provide context for searchers.
Metrics: Search metrics help companies understand what terms are most popular and where searches may be failing.
Multi-lingual: Separate search indexes may be stored for multi-language sites.
Relevance scoring: Relevance is a measure of how well a query matches results. Relevance scoring is used to determine the quality of the search algorithm.
Ranking: The order in which results are displayed — also called ranking — is critical to user experience and search success.
Visual search: There are many types of visual search technologies, such as image search, which might be used as part of a query in a browser or via AR/VR technologies.
Voice search: With the rise of personal assistants like Siri and Alexa, voice search has become one of the top mediums to perform queries.
Personalization: With enough data, results can be personalized using attributes such as gender, geo location, past search history, and more.
Recommendations: Recommendations can be driven through machine learning while leveraging the same search index.
Merchandising: Merchandising enables businesses to curate search results and browse content for specific promotions or campaigns.
A/B testing: Split testing the search UI and/or algorithm can help companies generate higher conversion rates and revenue.
Permissions: Assign users permissions to manage the search engine and results.
Hybrid retrieval: Keyword search is great for exact queries, and AI search is better for concepts or longer queries. However, you shouldn't have to pick one over the other. With hybrid retrieval, you get the best of both worlds.
Mobile search: Mobile browsing make up nearly 50% of online users for many sites. Search results and design needs to be adjusted for mobile usage.
Typo tolerance: Spelling errors can ruin even the bste seach engines. Ahem. Typo tolerance ensures it doesn't.

Use cases

Search and discovery support both vertical and horizontal use cases across different industries, screens, and devices.

Site search: Site search has been around almost as long as the the internet itself. These days, sites can include tens of thousands of web pages, files, and other records.
Federated search: Search across multiple sites and domains.
Browsing: Search engines can also power dynamic category and collections pages.
E-commerce search: E-commerce businesses have very complex requirements and need for deep integration across multiple technologies and devices.
AR/VR search: Queries in AR/VR may be driven by voice or visual search.
Marketplace search: With millions of SKUs and terabytes of data, marketplaces have tremendous complexity for search.
Enterprise search: Enterprise search often includes layers of security across federates systems.