When the site crashed, it posted an apology message for visitors: "We're sorry. Due to a system issue, you may be experiencing an interruption in customer service. The issue is affecting Discover's automated phone system, as well as Discover.com." Systems weren't restored until Friday about 5 PM New York time.
The apology also told customers that account information and related data "remain secure" and that their cards could still be used for purchases. (Now, now. No cracks about "Where?")
Discover itself opted to shed virtually no light on why the crash happened. But what Discover Senior PR Manager Laura Gingiss said is that call center reps accessed their customer-specific database (purchase activity and whatnot) through the same database that the site used. When the database crashed on Thursday, both the site and the reps were left data-less. "The reps couldn't access data, account information," she said in a phone interview. Call center reps confirmed, though, that they were able to take calls during the outage.
Gingiss E-mailed a statement that said—without explanation—the outage "had nothing to do with our infrastructure, network or telephony. Our VoIP [Voice-over-IP] network and telephony infrastructure were up and running without issue. Our toll-free number worked fine; cardmembers were able to call us and speak with customer service reps. We had a systems issue making some data temporarily inaccessible to our customer service functions."
The phone system issues are still confusing, though.The phone system issues are still confusing, though. Gingiss said the only phone issue was that phone calls to the call center were sometimes pointless because of the lack of data access. But that doesn't seem to gibe with the site's posted statement that "the issue is affecting Discover's automated phone system." That sounds more like the voicemail phonetree, which automatically transfers calls based on options selected from the phone, was down.
A frequent reason for sites and phone systems to crash simultaneously is when (VoIP) systems ride on the same network where customer databases reside. Gingiss confirmed that Discover is using VoIP but said "the interruption in our customer service last week was due to a systems issue unrelated to VoIP." Still, even if the crash was not "due" to VoIP, VoIP might still have been impacted by the outage and that impact might have manifested itself through phonetree problems.
Gingiss volunteered that the glitch didn't impact retail purchases. "Our POS systems continued to run without impact as well, so cardmembers were able to use their Discover Cards to make purchases," she said.
Another possibility—especially given the intense winter weather conditions in much of the U.S. during the time of the outage—could involve power interruptions. Gingiss said Discover does not outsource either its Web hosting or its call centers, which is unusual in and of itself, and these functions are in different locations.
The locations, though, may be irrelevant. If the database went out—for power or any other reason—it would have killed the data and wreaked havoc with the phone systems no matter where they were.
The timing of the incident's conclusion is a little unclear, too. Call center reps said everything was back up late Friday afternoon and Gingiss said it was early Friday afternoon. Time zone differences might explain part of that discrepancy. Another issue was when everything was fully restored.
Although "the technical issue was resolved sooner, the service interruption itself lasted longer to ensure that data was current and accessible by our customer service functions. The data was completely accessible by early afternoon Friday," Gingiss E-mailed.