THE SHORT ANSWER
An anomaly is a review signal, not proof of manipulation. Examine at least 90 days of follower and view trends, publication events, trends, collaborations, paid amplification, giveaways and platform distribution, then review repetitive comments, audience shifts and median comparable performance. Escalate only when multiple independent signals remain unexplained. Use outcomes such as explained, pending, unverifiable or does not meet requirements rather than unsupported labels.
Establish a normal fluctuation baseline
Record weekly followers, post volume and median comparable views for the same account. Platforms and formats grow differently, so one universal threshold is inappropriate.
Check whether change aligns with a breakout post, press coverage, collaboration or cross-platform traffic. When timing and audience context align, document an explainable event.
Combine at least three independent signals
Possible signals include follower spikes without view change, highly repetitive comments, sudden geography shifts, broken engagement patterns and concentrated new accounts. Each can have a legitimate explanation.
Review account growth, content performance and audience quality separately, then inspect their intersection. One weak comment sample should not override the full history.
Preserve evidence and counterevidence
Keep the anomaly window, source evidence, related content, team explanation and creator response. If the creator provides credible evidence of platform activity or paid amplification, include it in the conclusion.
The FTC treats followers, views, likes, saves, shares and comments as indicators of social influence and regulates fake-indicator transactions. Brands should review verifiable behavior without making public accusations.
Connect the review to a campaign decision
Low risk may proceed with post-launch data, medium risk may reduce scope or payment exposure, and high risk can pause for escalation. The action should match campaign value and brand risk.
Save review date and expiry in QuickKOL. New content three months later may change the decision, so one risk label should not become permanent.
Make anomaly detection an explainable review process
The FTC final rule identifies followers, views, likes, saves, shares and comments as indicators that can express social influence and regulates the sale or purchase of fake indicators. A brand should verify procurement and campaign risk rather than accuse a creator from one spike.
Split 90 days into 13 weekly windows and track net follower growth, post count and median comparable views. A week with 35% follower growth, 2% median-view growth and a 25-point geography shift can trigger review. A documented breakout covered by media may provide counterevidence. Set thresholds from account history rather than applying one rule across platforms.
Anomaly review table
Escalate on at least three independent signals and preserve explanations.
| Signal | Comparison | Explainable event |
|---|---|---|
| Follower change | Week over week vs 13-week baseline | Breakout or press |
| View distribution | Median, range, outlier | Paid amplification |
| Comment quality | Repetition, context, new accounts | Giveaway |
| Audience geography | Percentage-point shift | Cross-platform traffic |
CHECKLIST
Anomaly-data review order
This is an investigation order, not a fraud score.
- Verify evidence, date and window
- Build the account’s 90-day baseline
- Align content, press and paid events
- Compare median and range for similar posts
- Review comments and geography shifts
- Request minimal supporting evidence
- Record explanations and counterevidence
- Set expiry and next review date
Sources and further reading
- 01FTC: Final rule on consumer reviews, testimonials and fake social indicators
- 02TikTok Business: Top Content on TikTok One
External pages can change. Recheck platform rules, contracts and local law before launch.