The challenge
The challenge
More than 50 SharePoint site collections, created over a decade by different departments, could not be searched as one.
Approach
A decade of separate site collections
Over ten years, departments at the university had created more than 50 SharePoint site collections. Each had its own structure and each held information people needed, but enterprise search could not find it.
Students could not find course materials, and staff spent hours looking for policies. The university search returned thousands of irrelevant results, or nothing at all.
One index over every site collection
We crawled all of the site collections and captured document types, structure, usage and the relationships between documents, so the whole environment could be searched as one. Azure Functions keep crawling and indexing as content changes.
Search runs on Azure Cognitive Search, the service Microsoft now calls Azure AI Search, with AI enrichment tuned to the university's own terms and acronyms. A search for a course code also finds the course by its name.
Search where people already work
Instead of sending people to a new search portal, we built search into the places they already use, with SharePoint Framework components: the main intranet and the course management systems, as well as faculty portals and student dashboards.
Azure OpenAI interprets questions written in plain language and can answer them directly from several sources, instead of returning only a list of documents.
Tuned from how people search
Power BI dashboards show how well search works and where content is missing. Usage, such as which results people click and which queries return nothing, feeds back into the ranking.
Content authors see which of their pages perform well and what people search for without finding it. Later, students began to see resources matched to their courses, and faculty saw content related to their research.
Results
The outcome
All of the site collections can now be searched from the portals people already use. Orphaned content was found and archived, and authors can see the searches their content does not answer.
Stack
Technology used
- Azure Cognitive Search
- Azure Functions
- SharePoint Framework
- Azure OpenAI
- Power BI
The university is not named, as in the original write-up on our blog (12 May 2025).