Building a Repeatable Process for Improving Search

The Challenge

A national public health organization operated a search environment spanning multiple content collections, with information gathered through website crawling and database ingestion. Supporting that environment required both relevant results and dependable operations. The organization needed a consistent approach to search relevance evaluation so its team could assess proposed changes and track search quality. It also needed more efficient content updates considering indexing jobs associated with website crawling took more than a day to complete. Changes had to be prepared and deployed across a multi-collection search architecture that included disaster recovery. Improving search therefore required attention to evaluation, ingestion, index management, and deployment practices.

Innovent’s Approach

Innovent provided architecture guidance and hands-on support across relevance, content ingestion, and search operations. The work connected the quality of search results with the processes used to maintain the searchable content and deploy changes. The team combined relevance engineering with measurement, helping establish a repeatable approach to evaluating results against representative queries. This gave the organization a more consistent basis for assessing configuration changes than individual examples or subjective impressions alone. Innovent streamlined website crawling scripts to reuse existing crawl database data. Reusing that information reduced the work required for subsequent indexing jobs, shortening completion times from more than a day to approximately four hours. The team also addressed database ingestion and the processes used to update searchable content. Active-passive indexing allowed a replacement index to be prepared before switching search traffic to it. Alongside these improvements, Innovent provided search architecture guidance for the multi-collection environment and disaster recovery, and helped improve deployment processes. These efforts addressed how search changes were evaluated, prepared, and introduced into production.

The Result

Subsequent indexing jobs that previously took more than a day completed in approximately four hours after Innovent streamlined the crawling process to reuse existing crawl data. This shortened the time needed to prepare updated searchable content. The organization also gained a repeatable approach to search relevance evaluation, a process for preparing replacement indexes before switching traffic, and improved deployment practices. The engagement brought relevance engineering and operational improvements together, giving the organization a more structured foundation for maintaining search quality and managing change across its search environment.