Invoice Reconciliation Dashboard
Accountants were spending 6 hours per week manually matching invoices because the system didn't flag discrepancies.
The Problem
Relay's invoice reconciliation tool required accountants to manually match invoices to payments across multiple systems. The tool didn't flag discrepancies, so accountants spent hours hunting for mismatches that the system could have surfaced automatically.
- • Accountants spent 6 hours per week on manual invoice matching
- • Support received 150+ tickets monthly from users confused by the reconciliation flow
- • User retention was 68%, below the 80% target
- • Net Promoter Score was -12, indicating active dissatisfaction
The Process
I analyzed 6 months of support tickets and session recordings to understand where accountants got stuck. The pattern: they needed the system to do the matching, then review exceptions. I redesigned the flow as "auto-match, then review."
Support ticket & session analysis
Analyzed 6 months of support tickets and 200+ session recordings. Identified the "manual matching trap" — users spent hours comparing line items that the system could match automatically.
Auto-match with exception review
Redesigned the reconciliation flow: the system auto-matches invoices to payments with 95% confidence, then presents exceptions for human review. Accountants only touch the 5% that need judgment.
Confidence scoring & explainability
Added confidence scores to each match and explainability text ("Matched based on amount, date, and vendor ID"). Accountants could override matches and train the system on their decisions.
Gradual rollout & feedback loop
Rolled out to 50 accounts in phases. Created a feedback loop where accountants could flag false matches, and the system improved over time. Monitored accuracy and user satisfaction.
The Solution
We transformed the reconciliation tool from a manual matching interface to an intelligent auto-match system. The system matched invoices to payments with 95% confidence, then presented exceptions for human review. Accountants only reviewed the 5% that needed judgment.
- ✓ Manual matching time reduced from 6 hours to 45 minutes per week
- ✓ Support tickets related to reconciliation dropped by 41%
- ✓ User retention increased from 68% to 95%
- ✓ Net Promoter Score improved from -12 to +34
- ✓ The auto-match engine was licensed to 2 enterprise clients as a standalone feature