Web giants and the helpless individual

Like many users of technology today, I have developed an essentially dysfunctional approach when things don't work properly: I do whatever it takes to avoid fixing it. I wait to see if it 'fixes itself'. I make a workaround. I live with it till the next upgrade. Or I just use something else. It's only when I absolutely can't function without resolving the problem that I take a deep breath, grit my teeth, and embark on the quest to find a solution.

My worst nightmare is to find myself in the kind of situation frequently described in anguished blog posts by victims of Google, Amazon or eBay glitches and terminations. I've been collecting a few samples recently:

• When Amazon fails, it does so big time ranted fellow Enterprise Irregular Dennis Howlett last weekend after an ill-fated attempt to buy a high-spec digital camera from the online retailer. A security exception on his order triggered a defective process that closed his account with no viable means of getting it reopened.
• Nobody Can Hear You Scream If Your RSS Feed Is Dead wrote Louis Gray back in March when a glitch at Blogger wiped out his entire RSS feed, including all archives (among other casualties of the same glitch, Kent Newsome provided an entertaining and informative account).
• Google is Evil, Worse than PayPal: Don't use Google Checkout for your business wrote Amy Hoy later the same month when her website's Checkout account was disabled with no notification, no explanation and no appeal.
• Don't ever use Google Apps for anything important, wrote an anonymous poster to the Business of Software discussion community in January after being "stuck in this kafkaesque place" that is Google Apps support. [Updated 5 May at 22:52. An earlier version of this item incorrectly attributed the comment to Joel Spolsky, who is not having any problems with Google].
• Hello, Google, is anyone in there? I wrote last summer after several incidents when Googla Apps users were locked out of their accounts with no information.

As is the norm when these mass-market automated online services fail, the victims discover there's no information to be found anywhere as to what was the cause, when normal service would be restored, or what actions they can take to resolve the problem. At least for Gray and Newsome, a Google support advisor posted an explanation to the customer service discussion thread some 20 hours after the problem first appeared, and it was resolved later the same weekend. The Google Apps user solved his 'kafkaesque' domain problem by going direct to the registrar, bypassing Google. Amy Hoy got her money back, but only because she was able to make a big fuss that got noticed:

"It was only because I was angry enough to write about it publically, and that there was a community who supported & propagated that post, that I got this resolution. I have no doubt that if I just emailed Google, it would have gone ignored... I would have received empty form letters in response, and no action. Based on other people's experiences (just search for 'em), this seems to be the standard MO."

The common theme with all these stories is a fundamental flaw with the business model of cloud services, which is predicated on fully automated systems — fine when everything works as expected, but not fine when the failure is unexpected, unbudgeted or involves parameters the developer didn't think of when the system was designed. At least with cloud services, you can often hope that an operational problem will indeed 'fix itself', because the cloud provider may well be working behind the scenes to correct the fault. In that respect, it's better than when I have a recurring problem on my own PC, where the only resolution to expect is that it will cascade to a worse fault that I can't put off fixing (in which case I'll end up stuck in an automated support purgatory at Microsoft or HP's website). But if the cloud problem is a mission critical fault like a lost RSS feed or a failed payment service, then you can't afford to wait — and if the root cause is an account problem rather than something in the infrastructure, that's where the cloud model really falls down.

What these web giants need is an automated customer response system that acknowledges the possibility of exceptions. Instead of setting out to eliminate all human contact, they should explicitly allow for human interaction to investigate and resolve those problems that the system's designers haven't allowed for. Each problem resolved should then be analyzed to see how it can be eliminated by enhancing the automation — thus the human intervention becomes part of an iterative self-healing process through which the automation adapts to experience. It'll cost more in the short-term, but long-term, it'll enhance customer satisfaction and sales.

Another way to keep costs down is to do a better job of integrating online and community help systems — and being open about their capabilities and limitations. I know from my own experience that I'm often reluctant to investigate a problem online because I'm not familiar with the online process, which breeds mistrust. How long will it take to get a response? Will it answer my question? What do I do if it doesn't? I was impressed earlier this year with a briefing from community help platform provider Helpstream, which allows vendors to set business rules so that, for example, a question posed to the community can be converted into a case for resolution by an agent if it hasn't been answered within two hours, or if the originator isn't satisfied with the response. It is also working on processes that automatically monitor community respoonse and its effectiveness, for example by measuring satisfaction levels for specific pieces of advice.

More than any of these acts, though, the most constructive change would be to get rid of the mindset that leads these Web giants to belittle the circumstances of its 'consumers'. Is it unreasonable of us to expect to be notified if our account is being cancelled, or to want to know how long you think is acceptable for us to have to wait for a satisfactory answer to a support request? We are individuals — many of us with serious business dependencies relying on our usage of your services — and if you don't treat us with respect then sooner or later we'll take our patronage elsewhere.