Evan Schuman's StorefrontBacktalk

By now, almost everyone in retail knows about Amazon’s glitch that sent congratulatory victory E-mails to fans of both Alabama and Notre Dame—half a day before they played in the NCAA championship game on Monday night (Jan. 7). The obvious question: Why weren’t safeguards in place to prevent this from happening?

Amazon isn’t saying, but chances are good that such safeguards were in place—and that’s why the victory E-mails went out hours before the game even started.

After all, the best safeguard against any tech glitch in sending automatic victory messages as soon as the final gun sounds is to test in advance. Amazon undoubtedly wanted to make sure that the test was thorough, that its CRM system kicked out the full set of Alabama and Notre Dame customers, and that the E-mails were all generated and nothing coughed part way through. OK, that part of the testing went fine. The only piece of the morning test that failed was the part responsible for not actually sending the messages.

The irony is that Amazon probably was right to run the test and should have planned to send actual E-mails in the morning. It just shouldn’t have sent victory E-mails, instead sending some type of “good luck to (YOUR TEAM’S NAME HERE) in the BCS Championship!” message, complete with links to merchandise—just not actual BCS Championship merchandise.

Even smarter would have been to do that test a few days in advance, so customers would have had a chance to order game-day merchandise in time to have it in hand on game day. Then Amazon would have gotten two bites of that apple, in addition to having the test out of the way well in advance of the game.

That’s not the way IT shops are accustomed to thinking about testing, though. Tests aren’t supposed to be live. They’re supposed to be tightly controlled and tightly contained. That’s still a perfectly reasonable way to think about testing, especially when it includes anything involving payment or personal information.

But especially in a low-risk situation, the closer a test can be to real conditions, the better the test will be when it comes to turning up real-world problems. Triggering the test Notre Dame E-mails a day before the test Alabama E-mails, for example—and both of them days in advance of the game—would have provided good separation for test results in what was effectively a safe testbed and generated sales based on pre-game customer enthusiasm, as well.

(Face it: No.-1-ranked Notre Dame fans were definitely in a mood to buy team-branded products a few days before the game. Once the game was over, not so much.)

Yes, one safeguard failed. But Amazon missed a chance for a safer test and a selling opportunity. Isn’t that what No.-1-ranked Amazon is supposed to be good at?