Neiman Marcus Goes Down, But Only For A Special Few
Why are small problems sometimes the biggest pains? Sometimes because they're the hardest to spot. On January 25, Neiman Marcus' Web site was inaccessible only to customers using Internet Explorer versions 6 and 8 on Windows 7—everyone else was apparently able to get in without difficulty. This sort-of outage should have been easy to fix, but it lasted more than nine hours.
That suggests the Dallas-based high-end retailer made a change in the wee hours—exactly when you'd expect—but then accidentally left test code in the homepage. The result: a Web site that probably worked fine for everyone in IT, just not for all customers.
The odd outage, which was first reported by Internet Retailer, lasted from 3 AM to past noon on the East Coast. That's a long time, but walking through it is enlightening. At 2 AM Dallas time, new code went live—but presumably missing IE 6 and 8 from the complicated parsing most Web sites do to optimize each page for whatever Web browser the customer is using.
It wouldn't have been until maybe 7 AM on the East Coast that any customers would probably have tried the site with one of the browsers that didn't work. Any of those customers might have called Neiman Marcus' help line—there's a link to the number on the site's homepage. Except of course those customers couldn't get to the homepage to find the number to tell the retailer it had a problem.
(And most of those who tried another browser probably chalked the problem up to a weird glitch and didn't bother to call the company.)
It might have been hours before some customer finally noticed and decided to call (or someone at Neiman Marcus Direct spotted the problem). After that, it would have been the usual diagnose, patch, test and go live again. (That's all an informed guess; Neiman Marcus didn't respond to our request for enlightenment.)
Ironically, it probably took longer to fix the neimanmarcus.com problem because a well-designed E-Commerce site handles failures like this gracefully. Short of depending on fanatical customers to do your QA for you—or putting up a genuinely alarming error page that exhorts customers to call in the problem—the only practical alternative may be to much more closely monitor error pages after a site has undergone even trivial maintenance.
At least that will give you a heads-up when a whole class of customers goes missing—and you don't even know it's happening.