What does an 8.4% error rate cost a logistics company per year?
3.6M DKK — in lost billable services, credit notes, and customer complaints. A matching agent that reads freight documents (PDF and EDI), matches against WMS orders, and flags discrepancies brought the error rate from 8.4% to 0.9% and saved ~3.1M DKK/year. Billing delay went from 11 days to 2.
What was the problem?
The company handled ~45,000 shipments/month across 200+ e-commerce clients. Freight documents arrived in mixed formats — PDF, EDI, Excel — and had to be manually matched against orders in the WMS system.
One error meant: a credit note (direct loss), a customer complaint (support cost), and a billing delay (liquidity effect). At 8.4% error rate and 45,000 shipments/month, that was ~3,780 flawed shipments per month.
The estimated cost was 3.6M DKK/year — documented via credit note history and manual time tracking.
How is the system built?
- 1
Document reading
The agent reads incoming freight documents in PDF, EDI, and Excel via OCR + parsers. Format normalization happens automatically.
- 2
WMS matching
Normalized documents are matched against WMS orders via API. The agent calculates a matching score based on reference fields, amounts, and shipment data.
- 3
Discrepancy handling
Score above 90% and amount below 5,000 DKK: approved autonomously. Below 90% or above 5,000 DKK: flagged for human approval with explanation and diff.
- 4
Continuous operation
The agent matches continuously — not in batches. This reduced billing delay from 11 to 2 days and improved cash flow accordingly.
Concrete results (measured over 6 months)
- —Error rate: 8.4% → 0.9% — 89% reduction.
- —Credit notes: −78%.
- —Direct savings: ~3.1M DKK/year (documented via credit note history).
- —Billing delay: 11 days → 2 days — improved liquidity.
- —The agent is now primary processor for 96% of all freight document matching.
Questions about this case
- What is a typical error rate for manual freight document matching?
- Manual matching in logistics typically has error rates of 4-10%. At 45,000 shipments/month with 8.4% error rate, that is ~3,780 flawed shipments per month.
- When does the system require human approval?
- The agent escalates when the amount exceeds 5,000 DKK or the matching score is below 90%. All matches above 90% confidence within the threshold are processed autonomously with a full audit trail.
- How did cash flow improve?
- Billing delay went from 11 to 2 days because the agent matches continuously. That equals ~9 days of extra liquidity — at 3PL scale, a significant effect.