Financial & Enterprise
DAE Search
Making enterprise data actually findable
By Hiram BarskyFounder & Principal Designer
Role
Lead Product Designer
Timeframe
2024
Published
July 2026
Outcome
Retrieval time down 65% in user testing
Built with
GPT-4, Semantic Search AI · React, ElasticSearch, Python · Figma Design, Auto-Layout
The result
In prototype testing, analysts found the data they needed 65% faster — the number that anchored the business case for rollout. The quieter win: they could finally vouch for the results they acted on, instead of re-running analysis someone had already done.
Overview
An enterprise search platform where analysts couldn't find data that already existed. We redesigned it around the question they were really asking — can I trust this result enough to act on it. In testing, analysts found what they needed 65% faster.
The problem
Analysts were losing hours a day hunting for datasets buried across disconnected systems. The data existed; it just wasn't findable, and wasn't trustworthy once found. So decisions stalled while people re-ran analysis someone had already done.
Research
Discovery kept hitting the same wall: the data analysts needed already existed — they just couldn't find it, or trust it once they did.
Data silos were costing productivity.
Discovery barriers
Teams spend 3+ hours daily searching for existing data across disconnected systems.
→ Drove: Unified search interface with intelligent content tagging and federated results.
Permission complexity
Access control confusion leads to either data hoarding or security breaches.
→ Drove: Visual permission indicators and smart access request workflows.
Context loss
Found data lacks business context, making it unusable without tribal knowledge.
→ Drove: Rich metadata display with usage patterns and related content suggestions.
The approach
- 1
Replaced keyword matching with semantic search that reads intent, so a plain-language query returns the right dataset instead of a page of near-misses.
- 2
Made data lineage a first-class citizen — source, freshness, and transformation history on every result — so analysts could defend a number, not just find it.
- 3
Turned permission walls into guided paths: visual access indicators and one-click requests instead of dead ends.
Our thought process
Enterprise search isn't just finding files — it's understanding business context. We designed for the moment when someone needs to make a decision with incomplete information. The interface needed to bridge the gap between data discovery and business insight, making every search result a learning opportunity.
User testing
Prototype sessions with enterprise teams showed information retrieval time dropping to 5 minutes (vs 15+ previously), with search accuracy up 85%.
- 5 min
- Avg. retrieval time
- ↑85%
- Search accuracy
- 90%
- Found lineage valuable
What didn't work
Early versions tried to replicate consumer search patterns, but enterprise users needed more structure and context. A flat results list confused users who needed to understand data quality and permissions upfront. We also learned that auto-complete suggestions backfired when they exposed restricted content, creating security concerns.
The outcome
In prototype sessions, retrieval time dropped 65% — the number that anchored the business case for rollout. The real win was quieter: analysts acting on results they could finally vouch for, instead of re-running analysis someone had already done.
What this means for your project
Sitting on data, content, or internal tooling your own people can't find or trust? We redesign discovery around the real question — 'can I act on this?' — building search, provenance, and access for decisions, not just retrieval. It's the enterprise rigor behind everything else on this site.
“Unmatched in his ability to translate the often vague ideas from clients into beautiful, simple-to-use products.”
