Skip to content
Browse documentation

Third-party fraud scoring (IPQualityScore, Anura)

Score every new conversion with IPQualityScore or Anura by the IP address and user agent of its click, and reject or flag it automatically.

Last updated 4 min read

On this page

TrackCrown has its own click and conversion fraud checks (bots, datacenter IPs, proxies, minimum click-to-conversion time, conversion fraud rules). On top of these you can connect one third-party fraud vendor per workspace. The vendor scores each new conversion; high scores are rejected, medium scores are flagged for review.

Configure it under Tracking → Integrations → Fraud vendors (https://app.trackcrown.com/app/integrations?tab=fraud). Admins and owners can change it.

Supported vendors#

VendorYou enterScore used
IPQualityScore ("fraud score 0–100")API key (your IPQS private key)IPQS fraud_score, 0–100. Proxy, VPN, Tor and bot flags are kept as details.
Anura ("good / warning / bad")Anura instance ID (digits only)good = 0, warning = 80, bad = 100 (rule sets kept as details).

24metrics and other vendors are not supported.

Turn on scoring#

  1. Open Integrations → Fraud vendors.
  2. Choose the Vendor.
  3. Enter the API key (IPQualityScore) or the Anura instance ID. The key is stored encrypted; when editing, leave API key empty to keep it.
  4. Set the thresholds (1–100):
    • Reject at score — default 90. IPQS recommends 85–90. For Anura, 100 rejects only bad, 80 also rejects warning.
    • Flag at score — default 75. Must not be above the reject score ("The flag score must not be above the reject score.").
  5. Click Turn on scoring.

What happens to a conversion#

The worker runs every minute and scores, per workspace, up to 200 conversions per run that are:

  • created in the last 24 hours,
  • approved or pending (held conversions included),
  • not duplicates,
  • not scored before.

For each one it looks up the attributed click and sends the click's IP address and user agent to the vendor (IPQualityScore with strictness=1). The score is then compared with your thresholds:

ScoreVerdictEffect
≥ Reject at scorerejectStatus → rejected, status reason fraud:ipqs or fraud:anura, any hold is cleared. Partner postbacks that have not been sent yet are cancelled. The rejection is recorded in the audit log and the conversion's Change history, and fires the postbacks that opted in to rejected conversions; an approved conversion also fires its reversal postbacks once.
≥ Flag at scoreflagStatus unchanged; the verdict is stored on the conversion for review.
LowercleanVerdict stored, nothing changes.
No click IP (click-less conversion, e.g. coupon)skippedStored with detail "no click IP".

The verdict (vendor, score, verdict, details, time) is kept on the conversion. Rejected conversions are listed in Reporting → Conversions with the status reason shown in the conversion's detail; filter by status Rejected to review them.

Monitoring#

The card shows the vendor and key hint, and for the last 7 days:

FigureMeaning
Scored 7 dConversions with a stored verdict
RejectedVerdict reject
FlaggedVerdict flag

"Last scored …" shows the last successful call. Errors appear in red on the card:

  • Temporary errors (timeouts, vendor 5xx) — the conversion is retried on the next run.
  • Key errors (for example an invalid IPQS key) — the run stops for the workspace until you fix the key.

Conversions older than 24 hours are no longer picked up, so fix errors promptly.

Pause or remove#

  • Pause stops scoring new conversions; Resume continues with conversions of the last 24 hours that are not scored yet.
  • Remove — "New conversions are no longer scored. Past verdicts stay on the conversions."

Costs and privacy#

Each scored conversion is one API call to the vendor, billed by the vendor under your own account. The click's IP address and user agent are sent to the vendor; list the vendor as a sub-processor in your own privacy notice if you use it.

Something unclear or missing in this article?Tell us