Knowledge-base search without the ten-second wait

The problem

ESOMAR's ANA knowledge base gives more than 40,000 members access to over 10,000 pieces of content: journals, reports, webinars and videos across topics from big data to marketing and artificial intelligence.

Getting at it was the problem. After entering a search term, members sometimes waited ten seconds or more for a result.

Accuracy was the other half. A typing error often returned no results, or fewer relevant ones, with no correction offered. A knowledge base that big is only worth its membership if members can find what is in it, and a deteriorating search experience was a real risk to renewals.

What we built

dev-UP rebuilt the search on Azure Cognitive Search, a cloud search service with built-in AI that identifies and explores relevant content at scale and applies machine learning across different content types.

Semantic search understands what a member actually means by a query and ranks results by context rather than exact wording. Azure's Bing Spell Check corrects search terms as they are typed, so a typo no longer ends the search.

Advanced Search lets members narrow by date, author or company name. The solution reads the data sources ESOMAR already had through custom code, so no new data had to be acquired.

Access is restricted to members with the correct authorisation, and Azure's infrastructure scales automatically as the knowledge base grows.

The result

• Searches that once took ten seconds or more now return much faster
• Typing errors return corrected, relevant and properly ranked results instead of nothing
• Members can search by date, author or company name through Advanced Search
• The knowledge base stays open only to authorised members, on infrastructure that scales on its own

Related case

Every member question answered, at any hour