Master vendor data is critical for enterprise finance teams—poor data leads to fraud risk, duplicate payments, and costly errors. I redesigned the duplicate vendor review experience to help AP teams work faster with greater confidence.
[001] CONTEXT
Our platform flagged duplicate vendors, but the review experience was manual, slow, and delivered limited value. The legacy interface showed only basic metrics and required users to "Acknowledge" matches without seeing underlying data.
See duplicate flag
in Xelix
Click Acknowledge
No details visible
Leave to ERP
Context lost
Compare in SAP
Copy/paste IDs
Return to Xelix
Re-find record
Acknowledge
No feedback
Fragmented workflow with platform switching
[002] PROBLEM
Users had to leave the platform to verify vendors in their ERP, spending 3–4 hours daily with no feedback loop to improve detection. The workflow was fragmented and frustrating.
High False Positives
Rule-based fuzzy matching generated noise. ML not yet production-ready.
Example
False positive: Different entities
ERP Data Dependency
Vendor data pulled from SAP. Field availability varies by client configuration.
Data Flow
ML Training Required
Design must capture Yes/No/Ignore decisions to train future ML models.
Required Actions
Solution must work within these limitations while improving user experience
[003] CONSTRAINTS
Designs had to work with a rule-based system generating high false positives before ML was production-ready. The solution also needed to support a Yes/No/Ignore feedback loop to train future models.
[004] RESEARCH
Interviews with teams at Currys (£10B revenue), DS Smith (£7B revenue), and Liberty Global ($7B revenue) showed users trusted bank account and tax ID matches over names. "Having a lot of vendors open makes for a messier ledger. When you've got a massive master data file, you're not getting the benefits of scale. What you want is to have one good deal with one supplier." — Katie, Senior AP Manager
[005] DESIGN STUDIO
Facilitated a cross-functional session with PMs, engineers, designers, and the AI team to explore solution directions. This surfaced technical constraints early and aligned the team on a full-screen comparison approach with shared ownership.
[006] TESTING
Built and tested a prototype with customers based on design studio outputs and research insights. Users quickly recognised the value in improved workflow and duplicate detection, while some AP clerks highlighted the need for a more incremental rollout to avoid cognitive overload.
[007] SOLUTION
Redesigned the experience around side-by-side comparison, match visibility, and flow efficiency. Included full vendor comparison, match highlighting, pinning, progress indicators, and keyboard shortcuts to reduce friction and decision time.
[008] CUSTOMER IMPACT
Energy Transfer discovered 20% of vendor master data were duplicates, uncovering $1.7M in duplicate payments and recovering $540K within 60 days. Review time dropped from 3–4 hours to under 45 minutes (75% reduction).
[009] USER FEEDBACK
"We spend about 70% less time… everything we need is right there… what used to feel like a chore now feels manageable." — Jeff, Senior AP Manager, Energy Transfer
See match type
Bank, Tax ID visible
View Details
Opens overlay
Compare in Xelix
All data visible
Yes / No / Ignore
Keyboard shortcuts
Auto-advance
Trains ML model
75% time reduction — no platform switching
[010] SYSTEM IMPACT
The Yes/No/Ignore feedback loop now trains the ML model directly, turning a static workflow into a continuously improving detection system. Each user decision makes the platform smarter.