A specialist clinic: eight supplier catalogues reduced to one register
An orthopaedic implant clinic in the West Midlands was ordering from eight suppliers, each with its own catalogue conventions. Stock-outs were common and the surgical coordinator kept the real invento
The reading
A private orthopaedic clinic, two theatres, fourteen consulting surgeons on rotation. The clinic held consignment stock from eight implant manufacturers, each with its own part-number schema, its own expiry tracking, and its own idea of what constituted a lot.
The surgical coordinator, in post for nine years, had become the single human index. She knew that supplier A's 7.3mm screw was functionally equivalent to supplier C's 7.5mm in a particular revision. She knew which lots were approaching expiry and which reps were slow to replace them. None of this was written down.
When she took two weeks' leave in March, the clinic cancelled four procedures.
Eight catalogues, no shared language
The suppliers did not agree on how to describe an implant. One used material-first naming, another used anatomical-first, a third used a proprietary code that bore no relationship to anything else. The clinic's own paper register, kept in a ring binder in the sterile store, used a fourth convention invented in 2014.
The effect was that a surgeon asking for a particular plate had to be translated three times before anyone knew what was on the shelf. Reps brought in replacement stock that did not match what had been used. Expiry dates were tracked on a wall calendar.
- Eight supplier catalogues, four schemas, no shared identifier
- Expiry tracking on a wall calendar in the sterile store
- Average of 6 implant-equivalent lookups per procedure
- Lot numbers transcribed by hand onto theatre notes
- No electronic link between stock used and stock reordered
What we built
We built an internal register with a canonical schema: anatomical site, dimension, material, manufacturer, lot, expiry, consignment status. Each supplier's catalogue was mapped once, by the coordinator, over three afternoons. Thereafter new stock was scanned in against the barcode on the sterile pouch and the mapping handled the translation.
The register ran on a small Postgres instance in a London data centre, with a web front-end optimised for the theatre iPad. Barcode scanning used the iPad camera; we avoided dedicated hardware because the sterile store already had too many devices on charge.
The one meaningful technical decision was to store the full supplier payload alongside the canonical record. If a supplier changed their schema, which two of them did in the first year, we could re-map without losing history.
The numbers at ninety days
| What | Before | After |
|---|---|---|
| Implant lookups per procedure | 6 | 1 |
| Procedures cancelled for stock reasons | 4/quarter | 0/quarter |
| Expired stock written off | 3,100/month | 420/month |
| Reorder lead time | 11 days | 4 days |
| Coordinator hours on inventory | 22/week | 6/week |
The register is the memory
The coordinator is still in post, and still the most knowledgeable person in the building. But the clinic is no longer dependent on her being physically present for a theatre list to run. The register holds the memory now, and she holds the judgement.
That is the correct division of labour between a system and a person. We check it with the one-week training test: a new coordinator should be useful in a week. The clinic's new junior coordinator was useful in three days.