August 19, 2016
Guest Perspective: Peak Preparation
by Coby Montoya, Payments Industry Veteran
Having worked in e-commerce for 10 years now, I find myself using the word “peak” more often than “Christmas” or “the holidays” to describe the time between Thanksgiving and Christmas. Retailers—offline and online—make the bulk of their sales during this short time frame. Whenever I browse articles about e-commerce peak preparation, they tend to focus on the threat of fraud and reminding e-tailers not to let their guard down because the fraudsters are sneaking in with the good customers. I think this is the wrong message. Ecommerce companies need to do the opposite. Let your guard down but in a calculated way.
This is important for your sales as well as your brand reputation. Within the risk-management domain exists false positive. From the e-commerce risk manager’s perspective, this exists at the systemic level in fraud rules and the employee level in human error during manual review. More specifically, false positive is when a rule or human cancels a good order due to suspicion of fraud. Every fraud team has fielded calls from irate but legit customers demanding to know why their order was canceled. These escalations increase call-center contacts and slow down operations. If you leave your risk management process as is, you risk increasing lost sales dollars due to false positive by the same rate as your overall sales increase.
Less quantifiable but arguably just as important is brand reputation. Your brand reputation can be tarnished as angry customers vent their frustrations online. A single upset customer reaches hundreds of people in seconds though a Facebook update or Tweet. Hundreds can turn to thousands if friends share. This is why peak planning and preparation is so critical.
There are many ways to prepare but I find approaching from these three angles to be effective.
- Scale back fraud rules
- Expedite manual reviews
- Create bulk review and release process
Scaling back fraud rules means tweaking the risk dial in the opposite direction. If you use a risk engine that has a points system it means lowering the points of select rules. If you use an engine that has a hard decision (approve, review or deny) it means optimizing the conditions within select rules to reduce false positive rates. You should evaluate your top firing rules and determine which ones are contributing to a large percent of your manual review volume. It might be good to compare your average order value during non-peak vs. peak. If you notice an increase this might lead you to modifying order amount threshold rules. In addition to scaling back rules, try to identify historically good conditions and create positive rules that reduce the likelihood of manual review. Returning customer is an example.
Expediting manual reviews during peak is very important. The fraud review team should be looking for quick approvals. It can be argued that they should be doing this anyways, but this varies by merchant and risk level. It can be good to remind the team customers are innocent until proven guilty and not the other way around. The change during peak is spending less time to find traces of guilt. Merchants should be finding the fastest way to approve an order based on positive attributes. There is a risky, careless way to do this and an organized and methodical way to do this. It can be good to let a seasoned, top-performing manual-review agent define common, low-risk conditions. An example could be “paid email address domain name matches IP host name ISP name and order is less than X.”
Creating a bulk review and release process is a supplement to scaling back fraud rules. Due to product limitations, this is sometimes a necessary evil. What this entails is creating a script (usually in SQL) of moderate-to-low risk conditions that you do not have rules for. An SQL statement can be more targeted than a rule in some cases. Once the query is executed, the idea is to receive an output of order numbers for bulk release. Some vendors have functionality to perform this within the product. Others require the whole process be done on the backend through SQL statements. If this functionality does not exist in your fraud platform today it will likely take some times to build out, but it is highly effective. Once the low-risk orders are moved from the queue, the manual review team is left with orders that truly need a human eye.
I hope you find these tips helpful. Good luck during your peak season.