Your data doesn’t need to be clean
Free-text descriptions and accounting codes are the starting point, not a blocker.
Anonymized Use Case · Construction & Infrastructure
Your estimating problem is not an estimating problem
An estimate is a prediction of your own cost base. Most contractors and manufacturers have limited visibility on their contracts and spend. Illustrative data, based on a real pattern in construction procurement - not a named client engagement.
One item, ten descriptions, one price spread
One aluminium windowsill, bought on ten different projects, described ten different ways. Classified, the pattern is visible in days: here is what each project actually paid, and why two of them differ.
| Item (6 of 10 shown) | Supplier | Price |
|---|---|---|
| WS-ALU-2000 | AluBuild Ltd | £42.10 |
| Vensterbank alu | Vensterbank BV | £38.50 |
| Sill, window, aluminium | Sill Supplies UK | £51.20 |
| Windowsill alu 200cm | NorthFrame Aluminium | £44.00 |
| ALU SILL 2M | BuildRight Merchants | £36.75 |
| Aluminium sill 2000mm | Fenster & Co | £47.90 |
Aluminium windowsill, 200mm - UNSPSC 30161801 - 10 suppliers, 10 purchases. Price spread across the ten: £36.75 to £53.40, a 45% spread, on one item.
Two purchases sit above the median: one fast-track site with no lead time to shop around, one specified a heavier gauge for a coastal project. Both explainable, both invisible until classified.
Same scope, different price
A groundworks package, compared across 8 comparable projects, shows a 34% spread. Site access and a fast timeline explain most of it. Two quotes sit outside the band, each with a reason attached once classified. The other six show what the scope should really cost.
A code is not a category
A ledger item field like “6100-42-K” - structural steel fixings, M12 - is a nominal accounting code standing in for the item itself. Common in ERPs built for finance, not procurement. Nobody can benchmark spend they can’t see behind a code like that.
Pearstop works from invoices, ERP exports, purchase orders, and Excel - whatever format the data already arrives in - for cost-code benchmarking, outlier detection, and supplier consolidation.
AI-first, human reviewed
Every label checked. Classification holds up as new invoices and projects arrive.
Usage-based pricing
You pay for the lines processed. No heavy annual subscription.
Built and proven with a major Dutch infrastructure contractor, starting from zero classification history.
Typical result in specialised construction: 8-15% savings on addressable spend. The findings above are where it comes from.
Want to see this on your own spend?
Send a sample of your own procurement data and we'll show you the same kind of pattern in your own numbers.
