A fraud score is a numeric rating, often on a 0 to 100 scale, that estimates how likely a click, lead or conversion is to be fraudulent. It combines many individual risk signals into one number, so a program can apply consistent rules: accept low-risk traffic, flag medium-risk traffic for review, and block or reject high-risk traffic.
How it works
Each incoming click is checked against a set of signals, and each failed check adds weight to the score:
| Signal | What it catches |
|---|---|
| Data-center or hosting IP | Bots and proxies running on servers |
| Click velocity | Many clicks from one IP or device in a short time |
| User-agent and HTTP headers | Headless browsers, spoofed or malformed requests |
| IP reputation and blacklist | Sources already seen sending bad traffic |
| Device fingerprint | One device posing as many users |
The program then sets thresholds for each action. This is a trade-off: thresholds set too low block real users, while thresholds set too high let fraud through to payout.
Example
Example with illustrative weights: a click from a hosting-provider IP (+40) that is the 15th click from the same fingerprint in ten minutes (+30), with a normal browser user-agent (+0), scores 70. With a review threshold of 30 and a block threshold of 60, the click is blocked and logged.
Why it matters
A single check is easy to evade; a combined score is harder to game and gives affiliate managers a quick way to compare partners. A rising average fraud score for one affiliate or sub-ID can be an early warning that a partner has started buying low-quality traffic, often well before chargebacks or reversals show up. MassPartner scores every click using data-center IP detection, click-velocity limits, user-agent and header checks, IP reputation and device fingerprinting, as part of its fraud protection for brands. A broader checklist is in affiliate fraud detection.
Related: click fraud, cookie stuffing, self-referral, country blocking.