Author: Fraud Control Lead TG Target · filter stack · focus: anti-fraud entity, NOT moderation policy (moderation glossary).
The main thing from the article
- Anti-fraud = 3-layer filter stack.
- Click fraud vs fake FTD - different layers.
- Filtered events raise net FTD CPA.
- Report raw AND net CPA always.
- Crypto case: filter before scale saved €18k.
Fraud types taxonomy
| Fraud type | Signal | Filter layer | CPA impact |
|---|
| Click fraud | Bot/datacenter IPs, click spam | Pre-bid | Spend waste — no conversion |
| Bot registration | Velocity, device fingerprint | After click | Inflated reg, low reg→FTD |
| Fake FTD | Micro-deposit, KYC mismatch | Post-FTD | Inflated FTD, false CPA |
| Incentivized traffic | Reward pattern, geo anomaly | Post-FTD | Low LTV, high churn FTD |
Filter layers stack
| Layer | Timing | Action | Typical block rate |
|---|
| Pre-bid | Before impression | IP/device blocklist | 2–5% |
| After click | On registration | Bot score threshold | 8–15% |
| Post-FTD | After deposit event | Pattern + KYC validation | 5–12% |
Impact on reported CPA
| Metric | Formula | Includes fraud? | Use case |
|---|
| Raw FTD CPA | Spend ÷ raw FTD | Yes — all CRM FTD | Diagnostic only |
| Net FTD CPA | Spend ÷ net FTD (post-filter) | No — validated only | Scale decisions |
| Fraud rate | Filtered ÷ raw × 100% | N/A | Filter health KPI |
FTD CPA formula — FTD CPA glossary; CAC all-in — CAC glossary.
Impact anti-fraud filter methodology (TG Target, August 2026)
High risk accounts with vs without 3-layer filter, 90d rolling.
| Filter stack | Net FTD CPA variance | Post-scale fraud incidents/mo | LTV:CAC accuracy |
|---|
| 3-layer enforced | ±12% | 0.5 | ±9% |
| No filter | ±34% | 2.8 | ±31% |
Mini-case: crypto exchange EU - fraud filter before scale
Licensed exchange DE+AT · €60k/mo planned · FTD main KPI · 3-layer filter stack.
| Phase | Filter status | Raw FTD | Net FTD | Raw vs Net CPA | Decision |
|---|
| Pilot (no filter) | Disabled | 412 | 412 | €89 raw | Hold scale — suspicious reg→FTD |
| Post filter (Week 3) | 3-layer active | 398 | 341 | €89 raw → €104 net (+17%) | Scale on net CPA only |
Filter caught 14% fake FTD before €60k scale — raw CPA €89 looked on-target, net €104 revealed true unit economics. Avoided €18k waste on motivated traffic. Attribution intact — attribution glossary; postback match - postback glossary.
FAQ
What is anti-fraud in paid traffic?
Anti-fraud - filter stack entity for cutting off invalid traffic and fake conversions in paid acquisition: click fraud, bot registrations, fake FTD. Entity definition and KPI impact, NOT moderation policy. Moderation glossary - creative policy; here is the layers fraud filter.
Click fraud vs fake FTD?
Click fraud: Invalid/bot clicks that increase costs without human intent. Fake FTD: synthetic or incentivized first deposits passing by CRM, but not real users. Different layers of filters - before and after click - FTD. Taxonomy table below.
Filter layers — pre-bid, after click, post-FTD?
Pre-bid: Block known fraudulent IP addresses/devices before auction. Post click: bot score when registering landing. Post-FTD: analysis of deposit structure (speed, quantity, discrepancy KYC). Each level reduces reported conversions - affects the CPA denominator.
Anti-fraud impact on reported CPA?
Filtered conversions removed from FTD denominator → recorded increase in CPA compared to unprocessed ones. Example: raw FTD CPA 85 euros, post-fraud 112 euros (+32%). TG Target reports both raw and clean FTD CPA. FTD CPA glossary - formula; here is a fraud correction.
Anti-fraud vs attribution?
Attribution assigns credit to actual clicks. Fraud protection removes invalid events from the credit pool. Both affect the CPA message. Attribution glossary - window/model; antifraud - event validity filter.
Anti-fraud KPIs?
Fraud rate = filtered events ÷ unprocessed events × 100%. Click fraud rate (pre-bid blocking). Fake bid FTD (post-FTD rejects). Net CPA = expenses ÷ net FTD. <8% Fraud Target for EU Cryptocurrency and iGaming TG Target, August 2026
Anti-fraud and CAC?
CAC uses net FTD after fraud filter for unit economics truth. Raw CAC without fraud filter overstates efficiency. CAC glossary - all-in cost formula; anti-fraud adjusts conversion numerator.
Anti-fraud red flags?
Scale without filter post-FTD; FTD CPA suddenly drops (stimulated traffic); a sharp jump in geo-mismatch; reg→FTD heading anomaly. Action: Audit the filter stack before the budget ladder.
Anti-fraud benchmarks (Aug 2026)?
TG Target accounts with three-level anti-absence filter: net FTD CPA variance ±12% vs. ±34%; number of fraud cases after scaling: 0.5 per month versus 2.8 per month. Adoption of the filter for 71% of high-risk accounts in the EU.
Anti-fraud and reg-to-FTD?
Fake registrations increase the number of registrations; fake FTD overstates the number of FTD. Both distort reg→FTD CR. Reg-to-Glossary FTD - course formula; antifraud - validity filter by numerator/denominator.
Anti-fraud entity is current as of August 2026 (TG Target). FTD — Glossary FTD; reg→FTD — reg-to-Glossary FTD.