Roshanak Haghighi

Optimizing Operation of Fraud Report Review

  • My Role: UX Researcher
  • Team Size: 8 person
  • Completion Date: Mar 2023
  • Duration: 3 months
  • Status: Released

Non-disclosure agreement

Due to my NDA, I have avoided providing detailed information
The objective

Optimize fraud report review operation and reduce the operators’ review time by 20%

The Impact

Reduced the operators’ review time by 14%

Research Questions

Main Questions

  • What part should we optimize?
  • How should we measure optimization?

Sub Questions

  1. What are the pain points of fraud reports review operations?
  2. What are the top challenging parts?
  3. What is the average review time of the fraud reports review operations in total and at each stage?
Fraud report review operation process chart
Fraud report review operation process chart
Research Plan
Research Plan
Document the current operation process

Method 1: Contextual Inquiry with Supervisors

In order to clarify the current process, I started to gather information from the operators and documented around 300 operational tasks. Then broke down the fraud report’s reviews operations into 7 stages.

The confirmatory queue is similar to the primary queue but has a deeper review process in 7 stages

Here I created the list of details for each task and stage, as you can see in the table columns

  • Operational Task
  • Stages
  • Type of Operation: Doing, Switching between panels, Analyzing
  • Probability of Error: Low, High, Average
  • Place of operation: Panels and rule books
  • Does review time depends on the content of the ad being reported?
  • Judgement
  • Judgement Reason
  • Pain Points (Collected from Method 2)
Judgement Types

Method 2: Questionnaire from the primary and confirmatory review queues

Current Review Time Measurement

Data analysis of review action log

I analyzed the action log in Divar’s database to measure the current review time.

Review time of each stage

I created and ran a test with the review operators to measure the review time of each stage precisely.

Ideation based on pain points
Brainstorming @ Miro

Create an impact effort matrix to order the ideas

Clustered Ideas & Impact Effort Matrix @ Miro
Expected solution

Based on my research, the team developed an automated fake-detection system to recognize and tag each report in two levels Fake Report and Probably Fake Report, according to Divar’s existing fake-detection manual.

Result

Reduced the operators’ review time by 14%