What's interesting is not the percent drop (a 0.2 percent change is not likely statistically significant) but this claim in the news release accompanying the report's release: "Approximately one percent of accepted orders ultimately turn out to be fraudulent, but merchants also reject four percent of their incoming orders due to suspicion of fraud?some valid orders are almost certainly being left on the table."
According to Doug Schwegman, director of market intelligence for CyberSource, the one percent referenced (actually closer to 0.8 percent, he said) is just a theory, a guess from CyberSource. Although the surveyed retailers do reference a roughly one percent of fraudulent orders, the question asked about "accepted orders resulting in fraud losses," meaning that the one percent is on top of the roughly four percent of the orders that they have rejected because of suspected fraud. Schwegman said it's clear that some percent of the rejected orders are indeed legitimate and his team believes it to be about 0.8 percent.
It's an interesting issue because it's certainly easy to agree that some percent of rejected orders are actually legit and one percent seems as good a guess as any, but it's the precision of that guess (0.8 percent?) that starts to wander into the area of being too specific an answer for such an utterly unknowable concept.
With E-Mail SPAM filtering, it's somewhat easier to determine false positives because a consumers can look inside their SPAM folders and, over time, see roughly how many legit messages are caught for every 100 true SPAMs. (For my SPAM filter, it's probably about one in every 500. But given that I received almost 1,000 SPAM messages on a given day, that can mean as many as two legit readers messages a day that can disappear. Editor messages sometimes get lost that way, too, but I shed far fewer tears about that.)
But few retailers have such luxuries. Reviewing the list of rejected purchases each day does little good. A retailer theoretically could try contacting all of those people to see if any were real customers (and perhaps offering them some huge gift and an apology, if any are found), but I've yet to hear of one trying that. Most retailers rely on the much more passive approach of hoping that wrongfully wronged customers will call customer service and complain. It's more likely they'll go to the competition, badmouthing the first retailer all the way.
There's an even more important issue at play here. Let's assume for the moment that the 0.8 percent figure is precisely accurate. What is a retail IT manager supposed to do about that? Liberalize their fraud-detection rules? Not likely. First, getting a zero percent false positive system isn't going to happen.
With the acceptance of the inevitable errors, is slightly less than one percent such a bad figure? Put another way, how much more accurate is realistically achievable? Schwegman didn't have an easy answer, but argued that it would depend on the margins of a particular retailer. In the security-versus-profit balancing act, is slightly reducing the falsely-accused honest customers going to deliver more profit than will be lost by slightly increasing the number of actual frauds?
A simpler alternative for some retailers is to prominently post phone numbers and E-mails for senior customer service managers with any message sent to a rejected customer. Done properly, it would probably make life easy for the vast majority of incorrectly-rejected orders to be preserved after personal screening. Note: Those numbers had better not have huge hold times. Insulted, falsely-accused customers are not known for being especially patient.