A written decision instead of a judgement call
A branching decision flow, drawn in Word and shared as a PDF, so first line support reached the same answer on whether a case needed fraud review.
- Context
- Gaming platform, support and fraud operations
- Role
- Product Owner
- Date
- November 2017
The short version
- Result
- Processing time, errors and needless escalations down by an estimated 45%, at no build cost.
- What I did
- Mapped the fraud review decision into a branching flow, drawn in Word and shared as a clickable PDF.
- The call
- A document was enough, because the gap was a decision nobody had written down.
What changed
Support decided case by case whether something needed fraud review. The same case could get a different answer depending on who picked it up.
A short decision flow where a few questions lead to a yes or no on escalation. It was drawn in Word and shared as a clickable PDF so nobody could edit it by accident. Support still made the judgement, but the same case now got the same answer.
Problem
With money going in and out all day, first line support kept facing the same question, whether a case should go to fraud review. Each agent answered it their own way. The same case could be escalated by one person and closed by the next, which sent noise to the specialists and slowed everything down. The agents needed a decision path they could trust on a busy day, and that didn't need a platform project.
Approach
I mapped the financial and behavioural logic into a yes or no flow. The agent answers a few questions about the account, and the answers end at escalate or don't escalate. I drew it in Word and shared it as a clickable PDF, so agents could follow the checks without changing them by accident.
Support still made the call on each case, but agents stopped making it differently from each other. It cost nothing in software or development time.
Outcome
The PDF became part of the standard first line procedure. Processing time, errors and incorrect second line escalations came down by an estimated 45%, and the handover between support and fraud review got clearer. The 45% is my own estimate from being close to the work, and it wasn't audited.