Case Study · Knowledge Management
Enterprise Document Search Engine
A full-text search system over a large document archive, built to return relevant results in under a second at scale.
- Industry
- Knowledge Management
- Timeline
- 2025
- What we did
- Search infrastructure · Performance engineering · Technical SEO
The Challenge
What the client was up against
The client held a large and growing archive of documents that was, in practice, unsearchable. Finding anything meant already knowing where it lived. Naive database queries took many seconds and got worse with every document added, and none of the content was discoverable from a search engine.
Our Approach
How we solved it
Separate the store from the index
Documents and their files live in object storage and a document database; the search index is a separate, purpose-built layer. That split means the index can be tuned, re-analysed, or rebuilt from scratch without ever putting the source data at risk.
Tune for relevance, not just matching
We invested in the analysis pipeline — tokenisation, stemming, field weighting — so that the first result is usually the right one. Highlighted snippets show users why a document matched, which is what makes a result trustworthy rather than merely present.
Make the archive findable from outside
Result and document pages are server-rendered with proper metadata and structured data, so the archive is indexable by search engines. The content stopped being a private silo and started bringing in organic traffic.
What We Built
Inside the solution
- Full-text search with typo tolerance and stemming
- Faceted filtering across document metadata
- Highlighted result snippets showing match context
- Cloud object storage for source files with a separate search index
- Server-rendered, crawlable result and document pages
- Incremental indexing as new documents arrive
Technology
- Next.js
- Elasticsearch
- MongoDB
- Cloud object storage
- Node.js
The Outcome
Sub-second full-text search across the entire document archive.
- <1s
- Typical query response time
- Scales
- Index grows without degrading query latency
- Indexable
- Archive discoverable through organic search
Search is an infrastructure decision, not a feature checkbox. Bolting a wildcard query onto your primary database is a deadline you have borrowed against.
More Work
Other case studies
Different industries, same approach: understand the problem first, then build only what solves it.
Facing a Similar Problem?
Tell us what you're working on and we'll tell you how we'd approach it — no obligation.