- Anura backs its 99.999% BAD classification accuracy with a financial remedy, making its claim fundamentally different from competitors’ performance statistics.
- Competitors typically publish accuracy or false-positive rates, while Anura combines its accuracy claim with a guarantee and service-fee credit when a verified BAD classification falls below the guarantee.
- Anura uses multiple layers of verification and an “Innocent Until Proven Guilty” methodology designed to require strong evidence before definitively classifying a visitor as BAD.
- Rather than asking customers to determine their own fraud threshold, Anura makes a definitive BAD or GOOD decision when the evidence supports it and avoids classifying uncertain traffic as fraud.
Why Anura’s Accuracy Guarantee Separates it From Competitors
Plainly, Anura’s accuracy guarantee is different because of the fact that it is a guarantee. No other competitor puts their money where their mouth is for preventing false positives enough to guarantee their accuracy, and they certainly don’t offer refunds for their mistakes. Anura does, and that’s the difference.
Anura explicitly says its guarantee is 99.999% accuracy when identifying a visitor as BAD, using its Script integration. If a customer can provide verifiable evidence that Anura fell below the guarantee, Anura says it will investigate and, if validated, issue a service-fee credit for the affected month.
What’s Different vs Competitors
“We’re not asking you to trust our accuracy claim. We’re willing to put our money behind the accuracy of every visitor we definitively identify as fraud.” That’s different from saying: “Our model is 99.999% accurate.” For example, DataDome currently publishes 99.999% bot-detection accuracy with a false-positive rate below 0.01%. That’s a strong performance claim, but there isn’t a posted comparable accuracy guarantee with a financial remedy on the DataDome materials or website. CHEQ’s is the same. Its click-fraud product cites 99.2% accuracy versus 60–80% typical accuracy, while emphasizing risk-based decisions and low false positives. Its newer acquisition material claims a false-positive rate below 0.009% for fake-account prevention. But again, that’s a performance specification which is not the same as “if our BAD verdict is wrong, there’s a remedy.”
The other reason competitors can’t guarantee their claim is because they rely on scoring, not a definitive answer. Scoring models will look at a visitor, analyze the data collection and then give a numeric value to the fraud level and then force clients to make the final decision. An example may be a visitor receiving a score of 50% likelihood of fraud. That’s where the tool stops. It doesn’t inform you if that visitor is actually fraud. The decision is then left up to you, the client, on what your tolerance or threshold is, which can lead to that over-blocking or under-blocking. Set your tolerance too low, bad traffic get through and distorts your data or hurts your traffic quality. Set it too high, and real people can’t convert. Therein lies the difference. Competitors give you the best guess and let you decide, Anura gives you a binary answer to give you confidence on your decision.
How Can Anura Guarantee the 99.999% Accuracy?
This is less about the technology, and more about the team behind the technology. Anura’s CTO, Vince Kahn, describes the Anura method as “Innocent Until Proven Guilty”. “Most fraud tools are either too sensitive or too aggressive meaning they over-block or under-block. We’ve been in the industry long enough to understand that any real person has too much conversion value to make a mistake on, so we ensure that when we mark something as bad, we know that it’s bad because we double, triple, and quadruple verified, depending on the data we analyze, that the visitor is absolutely bad. This is why our environmental data and warning classifications are so valuable. Our bad traffic is always bad traffic, or your money back.”
Why a Warning Classification Matters
Not every visitor can, or should, be immediately classified as either GOOD or BAD. This is where a warning classification can provide an important advantage. Fraud detection systems operate in an environment where uncertainty is unavoidable. New devices, changing attack techniques, unusual user behavior, and incomplete data can all create signals that look suspicious without providing enough evidence to definitively identify fraud. A warning classification creates a third path between allowing a visitor through as definitively GOOD and blocking them as definitively BAD.
Instead of forcing uncertain traffic into a binary decision, a warning classification can communicate: “There are signals worth paying attention to, but there isn’t enough evidence to call this visitor fraud.” That distinction is important for preventing false positives. A legitimate visitor exhibiting unusual characteristics should not necessarily be treated the same way as a visitor whose behavior has been independently verified as fraudulent. This also complements Anura’s accuracy philosophy. If the goal is to guarantee the accuracy of a BAD classification, the system needs a way to distinguish suspicion from proof. Warning classifications provide that separation.
From Fraud Scores to Fraud Decisions
Fraud detection will always involve uncertainty. Fraudsters adapt. Devices change. Attack techniques evolve. The question is what a fraud platform does with that uncertainty. A scoring system passes the uncertainty to the customer in the form of a threshold: Where should we draw the line?
Anura’s approach is different: Do we have enough evidence to make the call?
Anura isn’t trying to predict which visitors are probably fraudulent. It’s designed to identify fraud when the evidence is strong enough to prove it and avoid turning uncertainty into a false positive. Learn more about your traffic with Anura’s free audit today so you can start making smarter traffic decisions.
