Every score, fully explainable.
No black box, no vendor magic. This page is the actual math and the actual rules — the same ones that run on your upload, described plainly enough that you could check our work.
The math
Average
Every rule that can run for a lead returns a 0–100 score. The final score is the average of those — nothing more. A rule that can't run (missing data, not enough upload history yet) is left out, not counted as zero.
<30 / 30–64 / 65+
Low, Medium and High risk cutoffs on that final score. We revisit them as accounts run more volume through the system, the same way any detection system gets refined over time.
Override
A match on your own blacklist sets a lead to High (100) outright, skipping the average — a value you've personally confirmed is bad beats any heuristic. An allowlist match does the same in reverse, forcing Low (0).
14 active rules
Built rule-by-rule around how affiliate fraud actually happens, not a generic bot-blocker repurposed for lead gen. Every plan runs the core rules below; Medium and High add the six that need this account's own cross-upload history — shared-IP rings, repeat offenders, payout outliers, interval regularity, relative affiliate benchmarking and offer-relative conversion timing. See plans and pricing.
Time-to-convert outlier
Flags a conversion recorded implausibly fast after the click for a person to have filled out the form, and separately flags one that's merely suspiciously fast. Skipped when either timestamp is missing — never guessed at.
Duplicate & near-duplicate detection
Catches the same email or phone appearing more than once — within the file you're uploading, and against every upload this account has ever run.
Duplicate lead, same affiliate & offer
A narrower, stronger version of duplicate detection: flags the same email or phone converting more than once for the same affiliate on the same offer.
Gibberish & disposable email
Flags addresses at known disposable-mail domains, and local parts that look machine-generated: mostly digits, no vowels, or one character repeated.
Your own blacklists & allowlists
Upload lists of known-bad emails, phones, IPs or names and any future match is caught automatically. Keep an allowlist too, so your best partners are never wrongly flagged.
Conversion velocity spike
Flags an hour where one affiliate's conversions spike well above their own normal pace elsewhere in the same file — a burst, not a trend.
Email pattern detection
Catches two addresses that resolve to the same inbox once dots and +tags are stripped, and groups of addresses on one domain sharing a base with a trailing counter.
Completeness score
Flags a lead missing fields — phone, IP, name — that the same affiliate normally fills in on their other leads in the file. A field an affiliate always sends is a field worth noticing when it's suddenly gone.
Conversion time, relative to this offer
Compares a lead's fill time against this specific offer's own leads, not a fixed threshold — a one-field quiz and a 15-field application have very different normal fill times, so a lead can clear the fixed check above and still be an outlier for its offer.
Relative affiliate benchmarking
Compares one affiliate's flagged rate on an offer against every other affiliate running that same offer — not a fixed global cutoff.
Shared-IP rings
Flags an IP address that several distinct email addresses convert from — a real household shares a wifi router; a ring shares a datacenter IP.
Interval regularity
Real people convert at irregular intervals. A script doesn't. Flags one affiliate's conversions landing suspiciously close to a fixed interval apart.
Repeat offender
Flags a lead whose email, phone or IP already scored High risk on an earlier upload for this account — fraud doesn't usually stop after one file.
Payout outlier
Flags a payout well above what's typical for its offer, computed excluding the lead itself so one outlier can't quietly pull up its own average and hide.
Fraud doesn't reset between files.
Duplicate detection and repeat offender don't just look at the file you're uploading right now — they check against every upload your account has ever run. A contact that scored High risk three uploads ago still counts today.
That history is scoped to your account alone: a database policy enforces it on every query, not just application code that could be forgotten in one route. Full detail on how data isolation works.
Every flag says why.
A flagged lead lists exactly which rules fired and the plain-English reason each gave — “Domain 'mailinator.com' is a known disposable-email provider,” not a score with no explanation attached.
A lead that wasn't flagged shows its own breakdown too, so “why wasn't this one caught” has a real answer instead of a shrug.
Nothing here is a black box.
There's no machine-learning model quietly deciding your leads' fate, no training set you can't see. Every rule on this page is logic you could read yourself, which means every score comes with a reason you can actually act on — dispute it, escalate it, or mark it confirmed fraud or confirmed legitimate for your own record. That marking never secretly changes the score behind it; what you see is what decided it.
Run your own file through it.
14 days free. See exactly which rules fire on your real data, and why.