A heritage builder: 220 years of material sourcing made searchable
A heritage building firm in the Cotswolds was sourcing specialist materials by phoning former colleagues. The historical knowledge of where to find a particular stone was held in two memories and a fi
The reading
A family heritage building firm, trading in the Cotswolds since the early nineteenth century, specialising in stone restoration and historic lime work. Thirty-two tradesmen, two directors, a sourcing manager in his sixties who was the fifth generation of his family in the firm.
The sourcing manager held, in his head, the location of every quarry, lime pit, and specialist timber yard the firm had worked with. He also knew which of them were still operating, which had changed hands, and which had a nephew now running them. The firm's younger staff, when faced with a sourcing problem, phoned him. He always had an answer.
He had mentioned retirement. The directors had, politely, asked him not to.
The memory that cannot retire
The filing cabinet in his office contained paper correspondence going back to the 1960s. Earlier records were in a locked chest at the main yard. There was no index. The sourcing manager could find any letter in a few minutes; no one else could find anything.
The firm's work was, by its nature, non-repeating. A seventeenth-century chimney needed stone from the same bed as the original; a Georgian cornice needed lime of a particular age and provenance. The matching of materials to buildings was knowledge, not reference.
- Approximately 180 active and historic suppliers
- Paper correspondence back to 1961, locked chest earlier
- No index of what stone had been sourced for which building
- All sourcing decisions routed through one person
- No written capture of why a particular supplier had been chosen
What we built
We did not try to digitise the chest. We built a supplier register with a specific structure: supplier, material type, geological or botanical provenance, historical use by the firm, current status, and (critically) a free-text field for the sourcing manager's own notes. Over four months, we sat with him twice a week while he dictated what he knew.
The technical choice: the free-text field was the point. We did not try to normalise his knowledge into categories. The structured fields handled the search; the free text held the judgement. A younger tradesman searching for "ashlar, Jurassic, local" would get a shortlist, each with the sourcing manager's own written note on why it was or was not suitable for a given context.
Stack: a small Rails application, Postgres with full-text search, a web interface that worked on the firm's iPads at site.
The numbers at ninety days
| What | Before | After |
|---|---|---|
| Suppliers in structured register | 0 | 174 |
| Sourcing decisions routed through one person | 100% | 40% |
| Historical building-to-material lookups | 2-3 days | <1 min |
| Younger staff making first-pass sourcing calls | 0% | 70% |
| Correspondence indexed and searchable | 0 | 1,100 letters |
The knowledge becomes transferable
The sourcing manager still works three days a week, by his own preference. The difference is that the firm now has a written account of what he knows, and the younger staff can make first-pass sourcing decisions without his involvement. He is used for the hard cases, which is the correct use of his time.
This is the one-week training test applied to a 220-year-old firm: a new tradesman should be able to make a reasonable sourcing call in a week. Not a brilliant one, not an experienced one, but a reasonable one. The register makes that possible. The man makes the brilliance possible.