Overview
The Fraud Score is a single number between 0 and 100 that summarises how likely a rewarded user is to be fraudulent. It is calculated automatically from the data collected across the seven Fraud Detection Reports and updated daily.
The higher the score, the stronger and more recent the evidence against the user. A brand-new user with no suspicious activity starts at 0.
ℹ️ For details on the reports that feed into the score → Rewarded: Fraud Detection Reports
Risk Tiers
Each score maps to a risk tier:
Tier | Score | What it means |
🟢 Low | 0 - 39 | No meaningful fraud signals, or only weak/old ones. |
🟠 Medium | 40 - 69 | One strong signal, or several weaker ones. Worth reviewing. |
🔴 High | 70 - 100 | Strong and/or repeated evidence of fraud. Recommended for action. |
Scores are conservative by design - a user only climbs into the Medium and High tiers when the evidence is clear.
How the Score is Calculated
The score is transparent and explainable - every point can be traced back to a signal you can inspect. It is not a black-box machine-learning model.
Each fraud signal contributes points based on four factors:
Severity - how serious the type of signal is. Several accounts sharing one device weighs more heavily than occasional VPN use, which is often legitimate.
Recurrence - how many times the behaviour was observed. Repeat occurrences add more points, but with diminishing returns.
Recency - how recent the behaviour is. Signals decay over time - as suspicious activity ages out of the lookback window, its contribution fades and the score drops on its own.
Correlation - related signals do not double-count. When several signals point at the same underlying cause, the strongest one counts fully and the others count partially.
The combined value is squashed onto the 0-100 scale, which saturates smoothly near 100 so the score can never exceed it.
ℹ️ Because the score is recomputed from a rolling window, it is self-correcting. A user who stops exhibiting suspicious behaviour will gradually return to a Low score without any manual intervention.
Where to Find the Score
Rewarded Users → Performance tab - a Fraud column shows each user's current tier and score. The column is sortable - click it to bring your highest-risk users to the top in one click. → Rewarded Users
Rewarded user detail → Fraud tab - the full picture for a single user:
A score gauge and risk tier
The list of active signals, each expandable to reveal its evidence and a View report link
A Recompute now button to re-run the analysis for that user on demand
If the user is banned for fraud, a banner with a Reactivate action
Rewarded Users → Payments tab - the Fraud Score is shown alongside payout requests so you can review risk before approving a withdrawal.
Fraud Policy Settings
The Fraud Policy page controls how the platform scans for fraud and how it reacts. Settings apply to your whole organisation.
Navigation path: Rewarded Users → Fraud Policy
Setting | Description |
User identifier | The identifier used to recognise the same rewarded user across all reports and the scan. Options: Rewarded User ID, Publisher Click ID, Unique 1-6. |
Daily scan enabled | When on, Swaarm runs the fraud reports every day and refreshes every user's score and signals. Turn it off to pause automated scoring. |
Lookback (days) | The rolling window the reports analyse (1-365 days, default 30). Older behaviour ages out of the score. |
Auto-ban enabled | When off, scoring only flags users. When on, users crossing the thresholds below are automatically banned. |
Auto-ban tier | Automatically ban users at or above this risk tier (e.g. High). |
Auto-ban score | Optionally, automatically ban users at or above this absolute score (0-100), regardless of tier. |
❗ Start with auto-ban off and let the platform flag users for a week or two. Review the High-risk users and their reports first, then enable auto-ban at the High tier once you are comfortable with the results.
Auto-ban and Reactivation
When auto-ban is enabled, any user who crosses your configured tier or score threshold is moved to Banned for fraud status and can no longer sign in to the rewarded app.
Auto-ban is designed to be safe:
It only acts on live accounts - it never resurrects an inactive user
It never un-bans a user automatically - even if a user's score later drops, you stay in control
To restore access, open the user's Fraud tab and click Reactivate. This returns the user to active status and their score continues to be tracked as normal.
Staying Informed
When a user is newly flagged as High risk, Swaarm can notify your team with an in-app and email notification. Configure who receives these alerts under Notifications settings. - ADD LINK when published - https://app.intercom.com/a/apps/fiz0w07o/knowledge-hub/all-content?activeContentId=18907389&activeContentType=article&editorMode=view&native_content=false
FAQ
Does a high VPN rate on its own get a user banned?
No. VPN usage is weighted lightly. On its own it will rarely push a user into the High tier - it usually needs to combine with other signals.
Why did a user's score go down without me doing anything?
Signals decay over time. As suspicious activity moves outside the lookback window, its contribution fades and the score naturally decreases.
Will turning off the daily scan delete existing scores?
No. Turning off the daily scan pauses recalculation. Existing scores and signals remain until the next scan or manual recompute.
Can I score a single user immediately?
Yes - open the user's Fraud tab and click Recompute now. This re-runs every report for that user with the current policy settings.