Read “The Global State of Fraud and Returns” report
Separating fraudsters from legitimate customers is a delicate business. Ecommerce merchants must protect their bottom line and provide stellar customer service that keeps customers coming back. So what happens when a transaction seems suspicious? Often, the answer is a manual fraud review.
But manual fraud review brings challenges. In fact, according to data from the Merchant Risk Council, 35% of fraud management teams cite “reducing or eliminating” manual fraud review as an area for improvement in their fraud management process.
In this blog, we’ll explain how manual review works, its drawbacks and benefits, and how ecommerce merchants can integrate comprehensive machine learning systems to prevent fraud and maximize revenue.
TL;DR
- Manual revue still has a role in ecommerce fraud protection. Human reviewers bring with them experience and intuition that machines can’t always match when faced with orders that are extreme outliers.
- Leading merchants today lead with a machine learning fraud protection approach, turning to manual review in the relatively rare cases that machines can’t instantly decision.
- Relying on machine learning fraud solutions for the vast majority of orders saves merchants money, increases their conversions and improves customer experience by all but eliminating false declines while speeding up fulfillment.
What is manual fraud review?
A manual fraud review refers to any time a transaction is analyzed for signs of fraud by a human rather than a machine. In the early days of ecommerce, manual fraud review was the main method of stopping fraudsters in their tracks.
Today, manual review still has a part to play in fraud prevention — but it’s often used alongside machine-automated review, and only if the transaction is complex or unusual. Data from the Merchant Risk Council’s 2025 Global Ecommerce Payments and Fraud Report shows merchants are now screening twice as many orders digitally as they are manually.
How do manual fraud reviews work?
Approaches to manual fraud review vary, but generally, they incorporate the same four steps:
- Initial assessment: This typically occurs when a specific fraud trigger or combination of triggers is detected (for example, orders over a certain value, unusually high orders from one cardholder or a mismatched shipping and billing address). Fraud teams review the transaction to determine if further investigation is required.
- Customer verification: Fraud teams review customer data, including address and banking information. They may also authenticate by SMS or phone, or contact the bank directly.
- Historical context: The team reviews past transaction data, purchasing history, and geographic location alongside customer data.
- Final determination: The individual or team makes a final judgment call on whether to let the order pass based on their findings. Typically, they create a case report including the details and outcome.
What is the cost of manual fraud review?
The cost of funding a fraud team varies from business to business. Some companies work with an in-house manual review team, while others outsource to specialized companies. Sometimes, customer service representatives double up as fraud prevention teams, which can also lead to extra payroll hours. According to Signifyd data, on average, a manual review costs $3.47 per transaction.
The costs of over-reliance on manual fraud reviews are far-reaching. They can lead to processing delays, which in turn lead to a higher abandonment rate — and a customer who feels inconvenienced is unlikely to return. Equally damaging, false positives caused by human error lead to lost revenue and disappointed, or even insulted, customers.
How does a manual review affect customer experience?
Sometimes, if a transaction is complex and a manual review is warranted, direct contact between a business and customer — when done well — brings a personal touch and a welcome extra layer of security.
Conversely, relying too heavily on manual reviews can negatively impact customer experience in several ways:
- Customer care: If a customer support team is under pressure to conduct manual fraud checks, customer service standards can slip.
- Brand damage: Nothing is more off-putting than having a transaction declined with no explanation. Or, worse, believing you’ve placed an order that never arrives. False positives cause fractures in customer relationships that often can’t be fixed.
- Order fulfilment bottlenecks: It takes time to sift through data and transaction history. Meanwhile, customers are left on the line, wondering when their card will go through or their package will arrive.
How effective is manual review vs. machine learning systems?
Both machine learning and manual reviews have a role to play in fraud prevention — but for the most efficient, watertight results, it’s best to let machine learning take the lead. The pros and cons of machine learning vs. manual review outlined below illustrate why:
Machine learning system advantages:
- Cost-effective: Machine learning systems can save merchants significantly on operational costs. Implementation costs are involved, but this is often far less than the costs of continuing with a fully manual review system.
- Efficient: Machine learning algorithms can detect fraud much faster than manual checks. This speeds up buying for the customer and saves time for the fraud prevention team to focus on complex cases.
- High accuracy: Machine learning models can access a wide range of data insights from a broad network. The larger the network of merchants, for instance, the greater the data insights and the more accurate the decision.
- Greater order fulfillment: A more accurate fraud review system means fewer orders declined by mistake.
- Scalable: Machine learning systems are infinitely scalable, which is invaluable during periods of high volume — fraud prevention teams won’t be left scrambling to pick up the slack.
Disadvantages:
- Lack of human nuance and context: In some more complex cases, machine learning lacks the human understanding and context necessary to determine if a transaction is legitimate.
Manual review advantages:
- Provides human judgment: For more complex cases of potential fraud, a manual check can bring human judgment and a higher level of contextual understanding.
- Extra protection: If a transaction is particularly high-value, a manual review can provide an extra layer of protection.
Disadvantages:
- Higher cost: When manual reviews are your primary method of fraud prevention, the associated costs quickly add up.
- Transaction friction: Manual reviews can slow down a time-sensitive process, leading to higher-than-necessary abandonment rates. According to data from Akami, even a one-second delay in load time increases abandonment rates by 7%.
- Hard to scale: Scaling a manual fraud detection system during high-volume times like the holidays is time-consuming and expensive.
- Subjective: A manual review process is only as effective as the fraud prevention team implementing it. In some cases, human subjectivity may lead to false positives and lost revenue.
- Limited data insights: With a fully manual review system, teams rely on internal data sets and publicly available records. As a result, data insights are limited and accuracy rates are lower.
Many businesses employ a hybrid approach to fraud detection, reserving manual reviews for only when it’s truly necessary.
Four ways to reduce reliance on manual fraud reviews
Manual reviews shouldn’t have to compete with the benefits of machine learning. Instead, ecommerce merchants can reduce reliance on a manual review strategy with the steps below:
1. Transform your fraud team into AI model managers
AI models bring extra layers of authentication — leading to reduced false declines, increased authorization rates, and a larger overall customer lifetime value. Thanks to its ever-expanding data set and continuous learning model, with platforms like Signifyd, merchants can approve up to 8% more orders.
2. Adopt machine learning that drives automated decisions
Integrating automated decision-making eliminates time-consuming, manual data analysis and decision-making.
But this doesn’t mean merchants need to relinquish control over their fraud prevention strategy — the most advanced solutions can be adjusted to align with individual risk preferences, affecting what percentage of transactions will be flagged for further review.
3. Leverage review-then-decline rates for optimal thresholds
Use a review-to-decline rate as a metric to determine if your automation threshold should be higher. Simply divide the total number of declined orders by the total number of manually reviewed orders:
For example: 100 transactions declined / 1,000 transactions reviewed = 10% review-then-decline rate.
A review-to-decline rate of less than 10% is a strong indicator to increase your automation threshold. Put another way, 90% of the 1,000 orders you reviewed could have been automatically approved, freeing your team up to focus on the more complex 10% that require their attention.
4. Transform your decision engine to lead with technology
When used as part of a wider automated process, manual review brings valuable insights. However, machine learning fraud technology works best when it’s the driving force behind your fraud prevention process—not just an assistant. This way, manual reviews can be reserved for when they’re most needed.
How to leverage AI and machine learning fraud protection
When considering the most advanced systems, machine learning solutions are able to accurately sort fraudulent orders from legitimate ones instantly, maximizing revenue and preserving a frictionless customer experience. For its part, Signifyd provides a comprehensive Commerce Protection Platform that protects the entire buying journey from account creation to returns and refunds.
Signifyd’s innovative machine learning solutions, the human intelligence of its data science and risk intelligence teams and its reliance on manual review in the small percentage of edge cases combine to create commerce protection suited to the 21st century.
Photo by Getty Images
Want to learn more about the leading edge of fraud protection? Let’s talk.
Frequently Asked Questions
What is a manual order review?
A manual order review refers to any time an online order is reviewed by a human for signs of fraud. Typically, manual reviews involve analyzing customer data, performing an authentication check, and reviewing transaction history.
Manual reviews are often used as one small part of a wider machine learning driven strategy.
What are the benefits and drawbacks of manual review?
In complex cases of suspected fraud, a manual review brings valuable, nuanced contextual understanding.
However, manual reviews are also expensive to conduct, and can lead to transaction delays or costly human error. During high-volume periods, like the holidays, a manual review process is challenging and costly to scale.
Manual review vs. automated systems: What is the difference?
Manual reviews are typically conducted by an in-house fraud team, outsourced to a company, or performed by customer service representatives. A manual review is always conducted by a human.
Automated systems can pull from large data sets to make instant, accurate decisions. Commerce protection partners like Signifyd use automated systems to provide guaranteed fraud protection, chargeback recovery and more.
How long does manual fraud review take?
According to the MRC Global Risk Council, it takes an average of 5.6 minutes to review a suspicious transaction.
