Blog / Machine learning & AI, Ravelin product

It’s what you do, not who you are: Account-focused fraud prevention unlocks optimal results

Transaction-based fraud defenses are largely outdated – but so are identity-centric solutions. Find out why an account-centric fraud solution delivers better results than legacy approaches while future-proofing ecommerce operations.

16 November 2026

It’s what you do, not who you are: Account-focused fraud prevention unlocks optimal results

Fraud prevention has evolved in leaps and bounds, with online merchants increasingly abandoning deterministic, rules-based approaches in favor of probabilistic, real-time monitoring. Ravelin has championed this AI-native strategy from day one, building our entire system on this premise.

However, there’s another dichotomy that is important to understand in fraud architecture, and it can make all the difference in the world.

Ravelin detects fraud by homing in on how each shopper account behaves – not on their real-world identity nor on isolated transactions.

Read on to discover the massive benefits of account-centric fraud detection compared to identity-based and transaction-based approaches.

What is account-centric fraud detection?

Account-centric fraud prevention treats the customer account as the persistent analytical object around which recommendations are calculated and provided.

In other words, account-centric fraud detection calculates fraud scoring and recommendations based on a holistic view of a shopper’s account signals and behavior instead of isolated shopper actions or a perceived shopper identity.

Account-centric fraud prevention protects ecommerce merchants across all stages of the shopping journey, exactly because it’s not dependent on individual transactions or identities. It goes beyond payment fraud to protect against first-party fraud, fake account creation, refund abuse, promo abuse, account takeover, fraud rings, and other fraudulent practices.

This type of fraud detection builds a continuously updated risk profile around a shopper/customer account for an ecommerce merchant.

Each new action is evaluated against the account’s history and established behavior, the merchant’s fraudscape, as well as any associated entities and consortium networks.

Account-focused vs identity-centric fraud detection

An identity based solution operates on the idea that knowing the real identity of the account operator is critical for fraud detection.

On the other hand, account-centric defenses such as Ravelin’s solutions operate on the idea that the behavior of the account holder in the context of the wider merchant behavior predicts fraud better than knowing who the operator is.

In other words, we believe it’s what a shopper does that makes them risky or not – not who they are.

Comparison table: Transaction vs identity vs account behavior fraud defenses

Transaction-baseIdentity-basedAccount behavior-based
Core analytical unitA standalone customer action as an isolated decision point – e.g. a purchase or a refund claim.The assumed identity of the person behind the transaction, often inferred across a merchant network.A specific shopper account on a specific merchant’s system as a persistent analytic object around which context and risk accumulate.
Primary focus on...Allowing or blocking specific actions.Allowing or blocking an identity based on that identity's record.The shopper's current behavior when compared to other customers' behavior and to this shopper account's historical behavior.
Core question asked“Does this transaction look fraudulent?”“Who is this person? Have they been lawful/legitimate in the past?”“Given everything we know about this account, this merchant, and this account’s network, how risky is the behavior they’re exhibiting right now?”
Scope & continuityNo continuity – each transaction/payment is scored individually.Global, cross-merchant data. Maintains a canonical profile across the internet dependent on real-life identities.Merchant-specific behavior and history based on what fraud vs legitimate behavior looks like for that merchant in particular.
Network link analysis capabilitiesLinks transactions to each other.Identity networks across all merchants in a vendor’s portfolio.Graph networks for each merchant identify linked accounts and catch multi-accounting, fraud links, suspicious behavior.
Key limitationsDisregards shopper history; cannot solve problems beyond payment fraud – or can do so with limited data.Stolen genuine identities/accounts & synthetic IDs can appear trustworthy. Privacy concerns & tools weaken matching and cause false positives/negatives. Insufficient to address all first-party fraud.Does not establish or verify real-world identities (unsuitable for those who need AML).
Ideal for...Small merchants with limited historical data. Merchants who require simple integration.Financial institutions who have AML requirements. Evaluating risk for brand-new accounts with no associated data or history.Enterprises. Companies looking to scale, detect fraud across the customer journey with respect to privacy laws/concerns.

Identity-centric fraud detection is not the answer: 7 reasons account-based detection is best

  1. Accounts provide more value: Defenses are exceptionally accurate and more agile, including against AI-powered fraud and abuse. Account-based approaches catch first-party abuse, synthetic identities and stolen accounts better.

  2. Saves time and money: Save yourself some work. As a merchant, you don’t really need to run KYC checks – in other words, you don’t need to spend time and money researching who the real-world person is behind an account.

  3. We live in a privacy-focused world: People have the right to be forgotten in many jurisdictions; legitimate buyers don’t always want to be known; large numbers of lawful consumers use privacy tools.

  4. Eliminates false positives from stale data: Consortium identities can be outdated. A phone number may be reissued; people change names; customers move house. On the contrary, Ravelin’s behavior-based defenses focus on real-time signals and ML features – behavior largely persists even when identity may show differences, such as a house move.

  5. Account-based is a more resilient way to detect fraud: If you make identity a tenet of your detection strategy, what happens when a regulator takes that away from you? And what about consumers using AI agents to shop – a legitimate method that hides their identities further?

  6. The best of all worlds: Building upon ecommerce fraud learnings of yesterday but looking firmly to the future, account-based fraud detection can still make use of the incremental benefits of identity signals and transaction signals without the false positives and false negatives associated with these older methods.

  7. More conversions: Ultimately, every ecommerce merchant cares about conversion, revenue and return custom. Account-based allows you to only block those accounts that do something wrong – not all of those who may share a name with a known fraudster, or once did something shady.

In short, Ravelin allows you to you block those accounts that did something wrong – not shoppers who share a name with a known fraudster, or customers who did something shady seven years ago.

How account behavior focused fraud detection works at Ravelin

Ravelin associates all activity with the merchant’s representation of a customer. We base our fraud detection primarily on how closely the customer’s behavior matches what we know about their historical behavior, as well as what we know about fraud at that merchant.

More simply: We don’t care who the customer is; we care about what they do.

The obvious fraud signals in fraud detection can be caught by almost every solution. But fraudsters continue to experiment and probe, now with AI as well. It’s the edge cases that bring the biggest losses to merchants, and those are best caught by looking at behavior.

There are fraud solution vendors out there who bang on about their ability to reveal the identity of a shopper – but what good does that do when a real customer decides to become an opportunistic refund abuser, or a criminal takes over the account of a legitimate shopper? What about a former fraudster who has changed their ways – do you want their custom or not? And those are just the obvious shortcomings; there are so many more.

Solutions that rely on identity may tar an account where the account holder is actually a victim, not a perpetrator, creating false positives that cause frustration, cart abandonment and loss of brand trust in legitimate consumers.

You need to be able to operate without identity data

In true alignment with privacy laws and best practices, the real-world identity of a customer does not matter to Ravelin. Even if it did, it’s a signal that we can’t guarantee we will have access to every region and forever. The direction of legislative travel is more privacy. In many regions, holding identity data and performing operations on it is already a contentious point and subject to strong regulation. As a company, you need to be able to operate without identity data.

Ravelin’s account-centric fraud detection identifies the techniques, methods and fraud signals employed by fraudsters and abusers attacking this particular merchant – including regional differences, gaps in defenses and loopholes in policies. This data-based approach utilizes AI fraud defenses, with multiple models deployed across the shopping journey to protect it end-to-end.

It should be noted that this approach still takes into account data points and machine learning features associated with individual transactions and user actions. However, these are all considered in conjunction with historical account data, for a clearer understanding of what is normal and what is an outlier for each account.

Over time, the account accumulates various points of history and in turn the AI model uses these to assess risk:

  • account creation and login history

  • successful and failed orders

  • transactions and payment methods

  • devices, IP addresses and locations

  • addresses, emails and phone numbers

  • session behavior

  • refunds, returns and chargebacks

  • manual reviews, comments and tags

  • links to other accounts and known fraud

What questions does account-centric fraud detection answer?

Because of its particular focus on user accounts instead of individual events across the shopper’s journey, account-centric fraud detection is best placed to answer questions such as:

  • Is this order consistent with the account’s history?

  • Is this account new or established?

  • Has the behavior suddenly changed?

  • What does this account normally buy and when?

  • Is this activity coming from a known device, address and location?

  • What is this account’s usual spend?

  • Is the card being used new or established?

  • Is this account profitable – what is its net spend after refunds?

  • Was this account previously tagged or reviewed as suspicious?

  • Is the account connected to accounts that were deemed risky?

At Ravelin, the answers to these questions are calculated as part of our machine learning AI features, and all together inform a fraud recommendation and a fraud score.

Moreover, although they come from AI, Ravelin’s fraud scoring is explainable: Clicking on it will show analysts the types of information that makes an account suspicious in the forms of families and megafamilies of features – for example, whether it’s IP addresses or bank cards that constitute the outlier.

Why scrutinizing behavior on account level works better across the customer lifecycle

The account creates longitudinal context. By collecting data at the customer level rather than the transaction level, Ravelin understands behavior and intentions.

This in turns allows us to prevent different types of fraud across the customer journey, including:

  • account creation

  • login

  • browsing and session activity

  • checkout and payment

  • fulfillment

  • refund requests and returns

  • later disputes and chargebacks

Use cases: Where account-based fraud defenses excel

Cohort behavioral fraud detection is exceptionally accurate. But there are certain use cases where it particularly excels, such as:

Stolen genuine identities

In 2026, social engineering is more successful than ever thanks to mass-market generative AI tools. An estimated 9 million identities are reported stolen every year worldwide, and in the USA alone, someone becomes a victim of identity theft every 4.9 seconds. One popular use for these is to conduct fraud against ecommerce companies.

Ecommerce enterprise fraud

Large-volume merchants are best equipped to take advantage of sophisticated account-focused behavioral fraud detection, as they have plenty of data, which increases the accuracy of the results.

First-party, “friendly” fraud

In the past year, 43% of merchants saw a rise in first-party fraud. Identity matters very little when it’s your own customers defrauding you – whether it be through false chargebacks, refund abuse, collusion with fraudsters, or promo abuse, to name just a few schemes. Plus, the same person can behave legitimately at one merchant and fraudulently at another.

Onboarding new accounts

The last thing you want to do is turn down new shoppers. By focusing on account behavior, you can safely accept payments from shoppers with little identity data – because they are young adults, for example, or after they’ve moved to a new country – instead of having identity-focused defenses block them or add too much friction to their experience.

Account takeover

A genuine customer with a real-world, confirmed identity creates an account. They shop with you for a while, until their account is taken over by a criminal, which they do not realize right away. The identity associated with the account is still valid and legitimate, but the account is no longer controlled by that identity – and the impact of account takeover can be detrimental to a merchant for exactly this reason.

Serving the unbanked and underbanked

For merchants and PSPs who support alternative payment methods, as well as guest checkout, setting up your defenses based on account behavior instead of identity is a no-brainer, as it allows you to serve more customers with confidence.

A note on identity data at Ravelin

At Ravelin, we still take into account identity data – as a factor of fraud detection, not as its goal. We do not base our decisions on a simple identity lookup. Identity attributes are inputs, not our governing theory. Where we have identity data, we make full use of it as inputs to our AI models in accordance with the law, together with a wide range of other fraud signals and AI model features.

Although we do make use of consortium data in our fraud recommendations, our popular graph network tool, Connect, is merchant-level; it shows you connections between customers that are unique to your company, and thus more meaningful than an identity graph. Connect also allows you to take quick action during a fraud attack, and even feeds features into our AI models.

An account-centric approach can significantly improve your fraud results

Merchants looking not just to stop fraud but minimize false positives and maximize conversions would be wise to use account-centric approaches to fraud defenses.

It’s true that your data could power any of these types of approaches – transaction-based, identity-based or account-based. But account-focused fraud prevention is the most efficient and provides you with a strong vantage point against contemporary online risk.

Ravelin Logo

Boost your successful conversions

Your bottom line deserves the most efficient, most advanced fraud defenses. Let's explore together what we can do for you.

Authors