What Users Want from Search at Enterprise Search & Discovery 2025

Relevance judgments are used to indicate which search results are relevant to a user’s information needs. Improving search is a circular, iterative process. The gather relevance judgments, measure search quality, generate hypotheses for improvement, experiment and validate improvement, evaluate and iterate, then start over.

Judging how good search results are depends on subject matter experts, expert users, LLM as judges, and implicit judgments. Each has its own pros and cons. At Northwestern Mutual, they formed a content strategy team to evaluate stakeholder happiness with the search experience. They started with reducing the cost of judgments and are looking to transition from lexical/keyword search to hybrid search, writing content for AEO and GEO.

Gurjeev Chadha, Simmplr, sees user expectations as a system that is fast, contextual, accurate, and continuously improving, yet finding information is still a struggle. Simmplr’s AI-powered employee experience platform is highly configurable. It has out of the box connectors and improves productivity with enterprise-wide across all connected systems directly within the intranet. It features a context and intelligence layer, a robust data and connectivity layer, and a secure and flexible platform. In terms of relevancy, it adds a knowledge graph with a learning to re-rank algorithm. Evaluation of search efficacy stems from feedback and measurement click-through rates.

KMWorld returned to the J.W. Marriott in Washington D.C. on November 17-20, with pre-conference workshops held on November 17.

KMWorld 2025 is a part of a unique program of five co-located conferences, which also includes Enterprise Search & Discovery, Enterprise AI World, Taxonomy Boot Camp, and Text Analytics Forum.