GED Finder — Turning an impossible task into a 3-second search
A solo project at Malakoff Humanis that turned a structurally impossible migration — find a buried PDF for each of 3,000 contracts across three slow network drives — into a sub-five-second lookup, used daily by the team.
Context
In early 2025 I joined Malakoff Humanis on a short-term contract to help process a migration of 3,000 insurance contracts into a SharePoint GED (gestion électronique de documents). Each contract had a corresponding PDF buried somewhere across three large network drives, and finding it was entirely manual.
The problem
Each contract required searching through 10,000+ subfolders on a slow network drive, using up to 12 different search terms per contract (contract number, folder code, SIREN, device code, etc.) plus cross-referencing two mapping spreadsheets. A single Windows Explorer search on a network drive could take 30 to 45 minutes.
| ~13h | 3,000 | 0 |
|---|---|---|
| estimated per contract, worst case | contracts to process | admin rights to install tools |
The task as designed was structurally impossible to complete. No indexing software could be installed without admin rights. The only available tools were Excel and read access to the network drives.
V1 — The insight
A Ctrl+F in Excel would always be faster than a Windows search on a network drive. If I could build an index of the drives inside Excel — folder names in column A, full paths in column B — I could search it instantly.
I wrote a VBA script to crawl the three network drives and populate that index. It worked: search time dropped from hours to seconds. But the solution only worked if you knew how to use it — which terms to try, which mapping sheets to check. It wasn’t usable by anyone else.
V2 — Designing for others
To share the tool with my team, I needed to abstract all the complexity. I built a UserForm UI directly in Excel — no installs, no admin rights, just a file anyone could open.
The core interaction: paste a contract number, and the tool automatically pulls every possible search term from the contract data and mapping sheets, searches all of them against the index simultaneously, and displays results ranked by hit count.
Key UX decisions
- Primary CTA: “Paste.” 90% of searches started with a copy-paste from the source spreadsheet. The affordance matched the actual workflow, not the theoretical one.
- Live search on keystroke. The input refreshes on each character and confirms “Exact match found” before proceeding — preventing wasted searches on the wrong contract line.
- Hit count per search term. Results show
(0)or(6)next to each term. Users skip zero-hit terms at a glance instead of clicking through 90 dead ends. - Stateful CTA. The primary button changes label and style once a match is confirmed, directing attention to the next step without instruction.
- “Ouvert” marker. A contract folder can appear in multiple search results under different terms. Once opened, it’s flagged to prevent duplicate uploads.
Outcome
Search time went from potentially several hours per contract to under 5 seconds. The tool was adopted by a colleague and was being rolled out to the wider team when the contract ended.
The project was never “in scope” — it started as a personal workaround to an unmanageable task, and became a product decision: if this tool has to work for someone else, every friction point I absorbed mentally has to be designed out.
Screenshots and live demo available on request.