Kount LogoBLog Against Fraud

Webinar Recap: How to Become a Manual Review Rock Star

posted on: Thu May 18 2017

If you missed the webinar “How to Become a Manual Review Rock Star,” this blog post provides a convenient summary recap. The webinar explored different tactics, techniques and strategies to streamline the manual review process while minimizing chargebacks and false positives.

Digital vs. Shippable Goods

Digital goods, such as e-tickets, software, music downloads, etc., share common characteristics that impact the manual review process:

  1. Consumer demand for instant gratification
  2. A highly commoditized market
  3. High volume, low cost goods, high margins (less overhead)

Because of these factors, digital goods merchants often have a higher tolerance for risk, which in turn impacts how manual reviews are conducted. In addition, the industry or items being sold affects the manual review process. For example, an order for 45 Beyoncé tickets on an e-ticket website will probably trigger a review, whereas an order for 45 seats of software on a B-to-B website probably won’t. Of course, there are many data points, automation and fine tuning that should go into making these decisions. The idea is to use Risk Scores to define which orders automatically get approved or declined, as well as which get manually reviewed. The goal is to dial in the Review Zone and optimize the revenue generated by reviewing and approving borderline orders against the fraud mitigation expense and chargeback costs of approving those orders.

Shippable goods, which can include automotive parts, apparel, computers, electronics, health and beauty items, home and kitchen products, etc., also share common characteristics that impact the manual review process:

  1. High ticket items
  2. High overhead or the cost of doing business
  3. Low volume orders, low margins

For these reasons, we often see a lower tolerance for risk. However, by implementing best practices, there is the opportunity to 1) be more aggressive and approve more orders and 2) use automation to lower the cost of manual reviews. Again, the goal is to find the point at which additional approved transactions generate more revenue than chargeback costs and product losses.

Prioritizing

Prioritization is always an effective tool, especially when you’re in a time crunch, short staffed, or ship internationally. For example, it makes sense to prioritize orders for review based on what time zones your warehouses or fulfillment centers are located. Orders that will be leaving your East Coast warehouse should be prioritized ahead of orders that will be shipping from your West Coast warehouse. Another way to prioritize is by whether or not an order is slated for next-day or 2-day shipping vs a slower shipping method. Finally, you can also prioritize reviews by time of order. In other words, orders that were placed first, get reviewed first.

Approve and Decline Lists

VIP lists streamline approvals and reduce manual reviews for customers with a good track record. These lists can be especially valuable for business-to-business transactions. Conversely, maintaining a list of known fraudsters and bad actors to be automatically declined can minimize or entirely eliminate manual reviews for those types of orders.

polling-biggest-challenge-manual-reviews.png

How to Tell a Story with Data

The starting point for identifying a fraudster includes checking the billing and shipping addresses, card information, and the order amount. Haly noted that if one of those three data points looks odd, she will check the shopping cart. If the items in the shopping cart are low risk, she will tend to approve. If the items in the shopping cart are high risk, she’ll investigate further. This can include looking at order history – has the customer ordered in the past, has orders shipped to the same shipping address, has that billing address been used in the past, etc. Haly cautioned that just because the recipient’s name and shipping address are different than the buyer’s name and billing address, that may not necessarily indicate fraud. The recipient could be a college student or a family member in the military. Other information to check includes device ID, the language of the browser, or other data points that indicate whether or not the order is originating from a legitimate customer. Ultimately, the best approach is to know your customers and understand what a good order looks like, so that the bad orders stand out.

False Positives and False Declines

It’s important to balance fraud reduction against revenue maximization. The dollars lost to false declines can be more detrimental than direct fraud losses. Javelin released a study last year that stated the dollars lost to false positives equal 19% of the total cost of fraud. The study also found that 25% of all declines for physical goods were false positives and 34% of declines for digital goods were false positives. What’s more, not only is the revenue from that one order lost, but there is the risk of losing the customer. In addition, word-of-mouth and social media can further amplify that negative customer experience and damage a company’s reputation. Because of these issues, Haly said her philosophy is: “when in doubt, ship it out.” The harm of offending good customers far outweighs the potential fraud loss. Laura remarked that identifying false positives and feeding them back into the system is incredibly valuable for helping improve approval rates. While merchants don’t always do this, it’s a great opportunity to boost sales.

polling-approval-rate.png

Effective Tools and Resources

Google, Facebook, LinkedIn, Spokeo, Talos Intelligence, and MarkMonitor are all good resources that are freely available.

  • Google: Search the web and quickly discover whether the person ordering actually exists and turn up other areas for investigation as well.
  • Facebook: Search to see if a person is legit – where they live, if they are a student traveling abroad, visiting a foreign country on vacation (different IP address vs billing vs. shipping address), etc.
  • LinkedIn: Does the buyer have a business profile, where are they located, do they split their time between the US/UK for work, etc.?
  • Spokeo: This database can provide name, address, employer, last 4 digits of phone, parent’s names, siblings, marital status, estimate of net worth, a photo of house, videos and photos, online profiles, and more for shoppers.
  • Talos Intelligence: This company produces known and emerging threat information for Cisco products, reports about spam and vulnerability, can search by IP, domain or network owner to see if there’s any relevant threat data.
  • MarkMonitor: This company provides brand and domain name protection which protects against internet counterfeiting, fraud, piracy, and cybersquatting, and publishes reports of brand abuse.

Paid tools like Chargebacks911, Ethoca, Experian, TeleSign, LexisNexis, Neustar, Emailage, etc. provide data useful for distinguishing good orders from bad orders, such as:

  • Chargeback data
  • Email lookup
  • Card issuer data
  • Identity data
  • Reverse phone lookup
  • Company lookup
  • Credit score
  • Friendly fraud data

Advanced Fraud Prevention Features

Online businesses, eCommerce operations, and other merchants processing card-not-present (CNP) transactions can minimize manual reviews, boost sales, and reduce fraud by capitalizing on advanced fraud prevention capabilities, including:

  • AI and Machine Learning: The processing power and memory capabilities of AI and machine learning far exceed human capacity for processing data and spotting trends in large data sets. Leveraging AI to manage transaction level details as a first line of defense can free up manual review agent resources. Further, Kount provides an AI rule recommender that analyzes 6 months of past data to produce future rule recommendations unique to your business.
  • Custom Weblinks: The idea here is that a click is faster than copy and paste. Having the ability to customize a direct integration with outside resources to further review an order can save hours per week per agent.
  • Automated Rules: The ability to create customized, automated rules can be used in multiple ways. It can be used to screen orders up front, based on parameters defined by you. It can also be used to route orders that need further screening to specific agents. This can be language-specific agents, or routing the highest value transactions to your most qualified agent.
  • Real-Time Decisions: Employing a fraud prevention solution that approves, declines or routes for further review within milliseconds is an industry standard. As consumers demand for immediate processing grows, this capability becomes even more critical for merchants.
  • Event Notification System: It’s likely that businesses are using more than one system for processing and shipping orders. Instead of clicking “approve” or “decline” in your fraud management system and again in your order management system, link these systems together so they are bi-directionally synced in real time.
  • Custom Reason Codes: Instead of manually typing or copying-and-pasting a reason to approve or decline a transaction, define the most common occurrences and pre-populate them in your fraud management system. This allows for consistency, accuracy, efficiency for your agents, and sharper reporting.
  • Order Linking: Kount call this indirect transaction association. It’s the ability to leverage the network effect of transaction data across hundreds if not thousands of merchants, showing how an order might connect to other orders using the same credit card, device, email, IP address and so forth.

Discover more about manual review best practices that can help you and your team review twice as many orders and improve accuracy by as much as 75%. Download the eBook “7 Best Practices to Become a Manual Review a Rock Star.” 

New Call-to-action

Topics
Share
TOP