Why businesses are getting fraud prevention wrong

For digital businesses confronting AI-assisted fraud, the pressure to tighten controls is considerable. Synthetic identities and generated documents challenge established verification processes, while automated attacks allow organised groups to exploit weaknesses at speed.

Know Your Customer (KYC) checks are a particular target, with experienced fraudsters using AI to mimic legitimate customers.

“Currently, KYC is the most targeted layer, where real-time document generation and deepfake face-swapping are becoming remarkably sophisticated,” says Volodymyr Todurov, chief executive of fraud prevention and risk management provider Frogo.

But the response can carry its own commercial cost. Aggressive controls may stop suspicious transactions while also rejecting legitimate payments, delaying withdrawals and driving customers elsewhere. A business can reduce its recorded fraud rate while losing valuable custom in the process.

Frogo, which combines device intelligence, automated risk scoring and real-time alerts, argues that businesses need to reconsider how they use fraud prevention.

“Operating with a ‘guilty until proven innocent’ mindset and relying on aggressive, automatic blocks,” Todurov warns. “Treating legitimate players like suspects completely destroys the customer experience.”

His argument sets a broader test for anti-fraud performance – whether it protects profitability once prevented losses, operating costs and the effect on legitimate customers are all taken into account.

Measuring the commercial impact

Todurov describes anti-fraud as a “revenue optimisation engine”, assessed through three connected measures – direct loss prevention, operational efficiency, and retention and lifetime value.

Direct loss prevention is the most familiar measure. It assesses how much money is being lost to fraud, including fraudulent withdrawals, and whether those losses are falling. But the resources required to achieve that reduction also matter. If routine checks demand constant manual attention, a business may struggle to manage growing customer volumes without continually expanding its risk team.

“Without a nuanced evaluation model, you risk blocking your highest-revenue players.”

Another measure is the number of monthly active users supported per analyst, alongside the time needed to complete investigations. Automating routine decisions should allow specialists to concentrate on more complex threats.

The third measure concerns what happens to legitimate customers. Todurov advocates “automated withdrawals built on the presumption of innocence”, linking smoother transactions to greater trust and long-term customer value. Delays and unnecessary checks carry a cost even when they never appear in a fraud report.

These measures need to be considered together. A system may correctly flag fraudulent activity while missing significant fraud elsewhere, or reduce losses through controls that also drive away legitimate customers. The financial outcome depends on understanding both what the system misses and what its interventions cost.

When suspicion becomes a costly mistake

High-value customers illustrate how easily aggressive controls can misfire. Their behaviour can differ substantially from that of the average user: frequent travel creates sudden changes in location, large deposits produce sharp spikes in activity, and multiple payment cards add further complexity.

Those same signals can resemble fraudulent activity, including attempts to test stolen cards. Without sufficient context, a business risks rejecting some of its most valuable customers.

“Without a nuanced evaluation model, you risk blocking your highest-revenue players,” Todurov says.

Frogo’s argument is that unusual behaviour should prompt an informed assessment. Depending on the circumstances, a signal may call for support or retention teams to intervene rather than an immediate block.

Recognising those distinctions requires specialists who understand the product and its customers, and can decide which signals justify intervention. That expertise also determines where AI can improve decisions and where established rules remain more effective.

“AI isn’t a silver bullet,” Todurov says. “Its success depends on specialists who know precisely where machine learning adds value and where tried-and-true, deterministic logic is still superior.”

Building a clearer picture from the first visit

Those specialists also need the right information on which to base their decisions. Frogo advocates gathering that information from a user’s first visit, beginning with the device they use to access the platform and developing the risk assessment as their activity unfolds.

“It all begins with device fingerprinting, where we analyse hardware and software parameters to verify user legitimacy and detect bots, spoofed devices, or masked origins,” Todurov says.

Frogo’s standalone Device Fingerprinting tool allows businesses to add this capability to an existing anti-fraud setup independently of its wider platform. Todurov identifies businesses with basic in-house rules but limited device recognition as a particular use case.

This adds context to signals such as a sudden change in location, helping analysts assess whether activity is consistent with a returning customer or warrants further investigation.

“Anti-fraud tools should be viewed as core data engines.”

The tool evaluates more than 300 client-side signals, according to Frogo, and produces identifiers serving two purposes: maintaining continuity within a single domain and recognising a device across brands within a portfolio.

It also supplies context such as IP risk, VPN and proxy information, browser specifications, screen resolution and hardware characteristics. Additional signals include time zone mismatches, remote desktop tools and virtual machine detection.

Within Frogo’s wider platform, information about an action such as registration or a transaction is checked against the business’s risk rules. Todurov says the system responds in under a second, explaining which risk triggers have been activated.

Turning fraud intelligence into business value

The value of that information extends beyond identifying threats. It can also help businesses recognise positive customer behaviour and act on it.

“Anti-fraud tools should be viewed as core data engines,” Todurov says. “Instead of only scoring bad behaviour, you can define triggers that identify high-value, ‘ideal’ player behaviour.”

Frogo’s real-time alerting system routes negative risk triggers to analysts while sending positive behavioural alerts to marketing and VIP teams through webhooks or Telegram. This allows different departments to respond to the same underlying activity according to their responsibilities.

Applying this approach in new markets requires continued adjustment. Customer behaviour, payment preferences and attack methods vary, so rules that worked in one setting may need revision elsewhere.

The commercial objective is to contain genuine threats, reduce unnecessary work and preserve straightforward journeys for legitimate customers as the business expands.

“The winners will combine a highly skilled fraud team with agile technology, viewing anti-fraud through a growth mindset,” Todurov says.