How to set fraud rules to protect your ad spend automatically
How to Set Fraud Rules to Protect Your Ad Spend Automatically
Fraud rules are conditional filters you configure in your tracker that evaluate every click against signals — IP type, click velocity, user agent, referer, uniqueness — and then block, flag or pass the click before it ever reaches your offer or gets billed by the network. Get them right and bad traffic stops costing you money without you babysitting a dashboard. Get them wrong and you block real buyers. This guide walks through the exact rule logic, thresholds and testing order for fraud rules affiliate tracking setups, using Adtraxo's rule engine as the working example.
What are fraud rules in affiliate tracking?
Fraud rules are automated conditions applied at click time. Each incoming click is scored against a set of signals; if it matches a rule, the tracker takes an action — block the redirect, flag the click for review, or pass it through but tag it. Rules run before the visitor lands on your page and before your postback fires, which is what makes them "automatic": you set them once and they work on every campaign, every hour, without manual review.
The distinction that matters: detection tells you what happened, rules decide what to do about it. Most trackers give you reporting. Fewer let you act on it at the click level. Adtraxo's Pro plan includes both fraud detection and rules, so a datacenter IP or a velocity spike can trigger an action rather than just appear in a chart.
Which signals should your fraud rules actually use?
Start with the five signals that catch the majority of junk traffic in performance campaigns. Each one maps to a specific rule type.
1. Datacenter IP
Residential and mobile users connect through ISPs. Bots, scrapers and click farms typically run from hosting ranges — AWS, DigitalOcean, Hetzner, OVH. A rule that blocks or flags known datacenter ASNs catches automated traffic that would otherwise burn budget on a push or pop campaign. This is the highest-value single rule for most affiliates. For the mechanics, see Datacenter IP detection: why it matters for affiliate tracking.
2. IP velocity
How many clicks from one IP in a time window. A real user clicks once, maybe twice. Ten clicks from the same IP in five minutes is either a bot or someone hammering your link. Set a threshold and block on breach. The trade-off is NAT and mobile carrier IPs, where many legitimate users share an address — covered in detail in IP velocity fraud explained — and how to stop it.
3. Bot user agent
Headless browsers, curl, python-requests, phantomjs and empty user agents are trivially identifiable. A rule that flags or blocks known bot UAs is cheap to run and rarely false-positives. The full list of tells is in Suspicious user agents: the hidden fraud signal in your click data.
4. Uniqueness conflict
Two clicks from the same IP with different device fingerprints, or the same fingerprint across many IPs, in a short window. This is the signature of a click farm rotating identities. Adtraxo's uniqueness conflict check flags exactly this pattern.
5. Invalid referer
Clicks arriving with no referer, a spoofed referer, or a referer that doesn't match the traffic source you're paying. If you bought PropellerAds pop traffic and the referer says a random forum, something is wrong.
How do you set a fraud rule step by step?
The process is the same regardless of which signal you're targeting. Work through it in order — one rule at a time, in monitor mode first.
- Pick one signal and one action. Don't build five rules at once. Start with datacenter IP → flag (not block).
- Set the threshold conservatively. For IP velocity, begin at something like 8–10 clicks per IP per hour. You can tighten later once you see the baseline.
- Run in monitor mode for 48–72 hours. Let the rule flag clicks without blocking. Export the flagged clicks and eyeball them: are these real users or bots?
- Check the false-positive rate against conversions. If any flagged click later converted (postback received), your threshold is too tight. Loosen it.
- Switch the action to block. Now the rule protects spend instead of just logging.
- Add the next rule. Repeat from step 1. Layer rules one at a time so you always know which one caused a change in performance.
- Review weekly. Fraud patterns shift. A rule that caught everything in month one may be stale by month three.
In Adtraxo, fraud rules live alongside the fraud detection dashboard under the campaign settings. Because Pro includes both detection and rules, you can watch a signal populate in the dashboard, then promote it to a blocking rule without switching tools. (Free plan has basic stats only — detection and rules are Pro features.)
What thresholds should you start with?
There is no universal number, and anyone who gives you one is guessing. What you can do is anchor to mechanism: the threshold should sit above what a genuine user could plausibly do and below what a bot needs to do to be profitable for the fraudster.
A worked example. You're running a sweepstakes offer on push traffic at a $0.40 payout. A genuine user clicks your link once, maybe re-clicks after a page error. A click farm needs volume to matter — dozens or hundreds of clicks per IP per day. So:
- IP velocity: flag at 5 clicks/IP/hour, block at 10. Well above real-user behaviour, well below farm volume.
- Datacenter IP: block outright. Legitimate push and pop users are essentially never on AWS.
- Bot UA: block outright. No false positives worth worrying about.
- Uniqueness conflict: flag at 3 conflicts per IP per hour, review manually before blocking — this one has more edge cases.
Run each threshold in monitor mode first, as above. Your actual baseline will differ by geo and vertical; a Tier-1 mobile campaign and a Tier-3 desktop campaign have completely different IP-sharing profiles.
How do fraud rules connect to sub-IDs and postbacks?
Rules decide whether a click is allowed through. Sub-IDs and postbacks tell you what happened after. Together they close the loop: you tag every click with a sub-ID, send it to the network, and when a conversion postback comes back you can see whether the converting click was flagged, passed, or blocked.
That's how you tune thresholds with evidence instead of vibes. If a click your velocity rule flagged later fires a conversion postback, the rule was wrong — loosen it. If a cluster of blocked clicks from one sub-ID never converts, the rule was right — tighten it. Adtraxo's sub-IDs (sub1–sub5) and S2S postback conversions are Pro features, which is the point: the fraud loop and the conversion loop live in the same tool.
The broader workflow for reading these signals is in How to read a fraud report and act on it.
Should you block or just flag?
Flag first, block second — unless the signal has near-zero false-positive risk.
Blocking is irreversible from the visitor's perspective: they hit your link and get nothing. If you block a real user, you've paid for the click and lost the conversion. If you flag a bot, you've paid for the click and learned something. So:
- Block immediately: datacenter IP, bot UA, empty UA, invalid referer. These are close to binary.
- Flag first, block after review: IP velocity, uniqueness conflict. Both have legitimate edge cases (carrier NAT, shared office IPs, families on one connection).
Once a flagged rule has run clean for a week — no conversions from flagged clicks — promote it to block. That's the whole discipline: evidence, then action.
How do fraud rules interact with your traffic source?
Different sources have different baseline noise. PropellerAds and EvaDav push/pop traffic carries more bot noise than native; Taboola and MGID are generally cleaner but more expensive per click; solo ads are a wildcard because you're buying someone else's list and can't see the source. Set rules per campaign, not globally, at least at first.
The practical sequence for a new campaign:
- Launch with detection on, rules in monitor mode, and a one-line arrival snippet on the landing page so you can see which clicks actually loaded the page in a real browser.
- After 48 hours, compare recorded clicks against arrival-snippet hits. A large gap means clicks aren't reaching a real browser — classic bot signature.
- Build rules from the signals that show up. Don't pre-build rules for fraud you haven't seen.
- Once rules are blocking cleanly, use A/B split testing to confirm performance improves — same offer, same budget, rules on vs off.
If you're also weighing whether a custom tracking domain changes how networks treat your traffic, that's covered separately in Custom domain tracking: does it reduce ad network fraud flags?
What do fraud rules cost you if you skip them?
You can't put a number on it without your own data, but the mechanism is simple: every fraudulent click is a click you paid the network for and will never convert. On a campaign spending $100/day where a meaningful share of clicks are bots, the loss compounds daily and silently — the dashboard just shows a bad ROI and you blame the offer.
Rules turn that silent leak into a blocked redirect. The cost of setting them up is an afternoon. The cost of not setting them up is whatever fraction of your spend is fake, every day, forever.
How does Adtraxo handle fraud rules compared to other trackers?
Adtraxo bundles fraud detection and rules into the Pro plan at $27/month (founding rate — it rises to $47 for new accounts; existing subscribers keep $27). That includes unlimited links, clicks and domains, A/B split testing, sub-IDs sub1–sub5, country analytics and S2S postback conversions, plus a 30-day money-back guarantee. The Free plan covers 10 links, 5,000 recorded clicks a month, one custom domain and basic stats — useful for testing the workflow, but detection and rules need Pro.
For context on the alternatives, as of October 2026 (prices change — check each vendor's page): Voluum's cheapest plan is "Profit" at $119/month billed annually; RedTrack's "Solo" plan is $169/month billed monthly or $118/month billed annually; ClickMagick's "Starter" is $79/month billed monthly or $66/month billed annually; BeMob has a free plan with paid plans from $49/month. Binom and Keitaro have their own licence pricing — check their pricing pages directly.
If you want the head-to-head on detection specifically, see Adtraxo fraud detection vs Voluum fraud detection — compared.
How often should you review and update fraud rules?
Weekly for the first month, then monthly. Fraud is adversarial: the moment a rule blocks a pattern, the operator behind it changes the pattern. A rule set that's perfect in week one is stale by week six.
A five-minute weekly review is enough:
- Open the fraud dashboard and sort by flagged clicks per campaign.
- Check whether any flagged click converted (postback received). If yes, loosen that rule.
- Check whether any blocked cluster shows zero conversions and high volume. If yes, that rule is working — leave it.
- Look for a new signal that's appeared since last week. Add one rule, monitor mode, repeat.
That's the whole loop. It's not glamorous, but it's what separates campaigns that scale from campaigns that quietly bleed.
If you're unsure how much of your traffic is actually fake, start there: How much of your affiliate traffic is fake? How to find out. And for the fundamentals behind all of this, The affiliate marketer's guide to click fraud detection and prevention and What is click fraud and how does it affect affiliate campaigns? are the right primers. For bot-specific signals, How to detect bot traffic in your affiliate campaigns goes deeper.
Frequently asked questions
What are fraud rules in affiliate tracking?
Fraud rules are automated conditions that evaluate every click against signals like IP type, click velocity, user agent and referer, then block, flag or pass the click before it reaches your offer. They run at click time, so bad traffic is stopped before it costs you a conversion or a network charge.
Should I block or flag fraudulent clicks?
Block signals with near-zero false-positive risk — datacenter IPs, bot user agents, invalid referers. Flag signals with legitimate edge cases — IP velocity and uniqueness conflicts — and promote them to block only after a week of clean monitoring with no conversions from flagged clicks.
What IP velocity threshold should I start with?
Start by flagging at around 5 clicks per IP per hour and blocking at 10, then adjust after 48–72 hours in monitor mode. Real users rarely click more than once or twice; click farms need volume. Your actual baseline depends on geo and device mix, so let your own data set the final number.
Do fraud rules work on the Adtraxo Free plan?
No. The Free plan includes 10 tracking links, 5,000 recorded clicks a month, one custom domain and basic stats. Fraud detection and rules, sub-IDs sub1–sub5, A/B split testing and S2S postback conversions are Pro features at $27/month (founding rate).
How do I know if my fraud rules are too aggressive?
Watch for flagged or blocked clicks that later fire a conversion postback. If a blocked click converts, the rule was wrong — loosen the threshold. Pair this with a one-line arrival snippet on your landing page to confirm that passed clicks are actually loading in a real browser.
Ready to stop paying for clicks that were never going to convert? Set up your first fraud rule in Adtraxo — start on the Free plan to test the workflow, then move to Pro when you're ready to block. Get started at adtraxo.com.
Track your affiliate campaigns for free
Clicks, conversions, fraud detection — all in one tracker.
Get started free →Not ready for an account? Get the Tracking Domination Playbook — free, just your email.