Hard services and FM companies manage thousands of invoice lines across hundreds of suppliers and sites. Pearstop classifies all of it automatically — 90–95% without human input — so your team can focus on category management, not data entry.
FM companies face three procurement data problems that most tools are not built for.
Invoice lines like 'elektra werkzaamheden Q3' or 'pump seal replacement unit 4B' are meaningful on-site but impossible to classify consistently at scale. The same physical work appears under dozens of different strings across sites.
A large FM operation may buy from 500–3,000 suppliers. Each invoices differently. Without a consistent classification layer, spend data cannot be aggregated or compared — making category management guesswork.
At 5,000–35,000 invoice lines per month, manual UNSPSC classification requires one or two dedicated staff working continuously just to stay current.
Pearstop's engine handles 90–95% of FM spend lines automatically — and gets better the longer you use it.
Supplier-specific and GL-specific rules handle high-volume, high-confidence classifications immediately.
An ML layer trained on your spend history replicates the classification logic your best category managers apply.
Ambiguous descriptions, foreign-language lines, and edge cases are resolved by a large language model with deep product and industry knowledge.
Items below the confidence threshold are flagged for your team. Every decision feeds back into the engine.
90–95% of invoice lines classified without human input after the initial learning period.
Processing 35,000 procurement lines monthly for a major Dutch infrastructure contractor.
FM teams typically reduce manual data work by 70–90%, freeing buyers for category strategy.
Paste any product or service description and get the correct 8-digit code instantly — or explore the full UNSPSC tree to understand the hierarchy.
Our asset lists worked for mechanics on-site, but did not allow us to plan smart maintenance or manage bid risk in a data-driven way.
A facilities management company typically buys from 500 to 3,000 suppliers across every site it services, and each one invoices differently. UNSPSC classification assigns the same 8-digit commodity code to the same underlying work, whether an invoice line reads 'elektra werkzaamheden Q3' or 'electrical maintenance callout', so procurement and finance can compare spend across suppliers and sites instead of reading free text one line at a time. Pearstop's engine does this automatically for FM companies processing 5,000 to 35,000 invoice lines a month, classifying 90 to 95% of them without a person touching the line.
The most common segments for FM are Segment 72 (Construction and Maintenance Services), Segment 76 (Industrial Cleaning Services), Segment 80 (Management and Administrative Services), and Segment 73 (Industrial Production). Hard services FM companies typically see the highest spend volume in Segment 72, covering electrical, HVAC, plumbing, and fabric maintenance.
FM companies buy from hundreds of suppliers across decentralised sites. Invoice lines are often free-text descriptions written by field engineers, not procurement teams. Manual classification at this scale requires one or two dedicated staff working continuously just to keep pace.
An initial classification of a full spend dataset typically completes within four to six weeks, including a Data Stability Baseline to validate accuracy. Ongoing monthly classification runs automatically.
Pearstop receives data via CSV export or direct API connection and returns classified data in the same format, compatible with any CAFM or CMMS system including IBM Maximo, ServiceNow, Planon, and SAP PM.
Pearstop's four-layer engine achieves 90–95% automatic classification on typical FM spend datasets. The remaining 5–10% is flagged for human review, and each reviewed decision feeds back into the engine.
Explore the full UNSPSC classification service, real FM case studies, and the broader procurement data quality offering.
Talk to sales. We will show you exactly how the engine works with your spend data.
