Analyzing abuse in the content‑reporting system

Quick Overview

Evaluates reporting-system analytics for valid reports and misuse detection in moderation workflows. Strong answers define distinct reported-user share, write SQL for valid reports, and detect abusive reporters through frequency, hit-rate, burst, and coordination metrics.

Analyzing abuse in the content‑reporting system

Company: Meta

Role: Data Scientist

Category: Analytics & Experimentation

Difficulty: medium

Interview Round: Technical Screen

Scenario: Content‑reporting tools can be gamed. Calculate valid‑report ratios and detect users spamming false reports using SQL‑friendly metrics. ​ Question 1: Compute the share of reported users with at least one valid report. (Hint: numerator/denominator definition, deduping) Question 2: How would you detect misuse of the reporting function? Provide metrics and SQL. (Hint: report frequency vs hit‑rate)

Quick Answer: Evaluates reporting-system analytics for valid reports and misuse detection in moderation workflows. Strong answers define distinct reported-user share, write SQL for valid reports, and detect abusive reporters through frequency, hit-rate, burst, and coordination metrics.

|Home/Analytics & Experimentation/Meta
Meta logo
Meta
Jul 12, 2025, 6:59 PM
mediumData ScientistTechnical ScreenAnalytics & Experimentation
12
0

Measuring Valid Reports and Detecting Abuse in a Reporting System

Analyze a user reporting system over a 30-day window. The schema is:

reports(report_id, reporter_user_id, reported_user_id, content_id, created_at, is_reviewed, is_valid)

is_valid is defined only when moderation review is complete. Compute the share of reported users with at least one valid report and detect misuse of the reporting function.

Constraints & Assumptions

  • Deduplicate reported users for the share metric.
  • Pending reviews do not count as valid.
  • Reporting misuse includes frequent low-hit-rate reporting and coordinated or spammy reporting.
  • SQL should be clear about numerator, denominator, and time window.

Clarifying Questions to Ask Guidance

  • Should the denominator include users with only pending reports?
  • Are repeated reports for the same content or user deduped?
  • Is is_valid reliable, and how long does review take?
  • What action will be taken against abusive reporters?

Part 1 - Share of Reported Users with a Valid Report

Compute the share of reported users who have at least one valid report in the period.

What This Part Should Cover Guidance

  • Denominator: distinct reported users with at least one report in the window.
  • Numerator: distinct reported users with at least one reviewed and valid report.
  • Handle pending reviews explicitly.
  • Provide SQL with grouping by reported user.

Part 2 - Misuse Metrics

How would you detect misuse of the reporting function?

What This Part Should Cover Guidance

  • Compute reporter volume, valid-rate or hit-rate, invalid-rate, unique reported users, repeated content reports, and burstiness.
  • Compare reporters against peer baselines and require minimum volume before flagging.
  • Identify coordinated reporting and target concentration.
  • Include SQL focused on report frequency versus hit rate.

Part 3 - Follow-up Actions

What would you do with suspected misuse signals?

What This Part Should Cover Guidance

  • Use thresholds, review queues, warnings, rate limits, credibility scores, or product friction.
  • Avoid suppressing legitimate high-volume reporters.
  • Monitor appeal outcomes, moderation accuracy, and safety impact.

Follow-up Questions Guidance

  • How would review delay bias the valid-report share?
  • How would you detect coordinated brigading?
  • What if a reporter has low hit-rate but reports genuinely hard-to-review content?
Loading comments...