Ronny Richardson tells the tale, which begins with this e-mail to the Walmart.com Help Desk:
I sent a money gram to Cameroon. The local clerk said they could deliver to Yaounde Cameroon and the person I was sending it to could pick up the money at the local Walmart. As far as I can tell, there is no Walmart in Yaounde. Where would they go to pick up the money?
He received this response, which I’ve edited for length:
Dear Ronny,
Thank you for your inquiry regarding Walmarts MoneyCenters. At Walmart, we believe everyone should have access to affordable financial services, blah, blah, blah, blah, blah.
We have been offering basic money services in our stores since 1999, typically at the customer service desk, with the first MoneyCenter opening in 2004. We currently have 225 MoneyCenters today and will open another 225 before the end of this year. By the end of 2008, yammer, yammer, yammer, yammer, yammer.
In early 2007, Walmart redesigned the MoneyCenter format enhancing the customer experience by burble, burble, burble, burble, burble.
Again, thank you for your interest in MoneyCenters and we hope you continue shopping at Walmart.
Sincerely,
Walmart Customer Care
As Mr. Richardson points out, “Three things are very striking about this reply:
1. It is clearly a canned response.
2. It clearly does not answer my question or even really touch too closely on it.
3. It was clearly written in or before 2008 and never updated!”
It’s easy to sneer at a misfire like this. But while sneering at Walmart is gratifying, it doesn’t put money in your pocket.
Understanding how an organization of Walmart’s size and sophistication fell into this trap so you can avoid the same pitfalls? That pays the bills.
As is usually the case, I’m unencumbered by facts, which leaves me free to draw whatever inferences I please, so long as they’re plausible, such as:
Pitfall 1: Default optimization
The default optimization is cost reduction.
When modifying a business process or practice, or designing a new one, whoever is responsible should clearly and consciously rank the six dimensions of process optimization (which, if you haven’t read the KJR Manifesto or Bare Bones Change Management, are fixed cost, incremental cost, cycle time, throughput, quality, and excellence).
If they don’t, businesses inevitably minimize net cost, thinking terms like “return on investment” (ROI) make them sound sophisticated instead of unconscious.
When companies go for cost-reduction the unconscious way, they end up reducing quality and excellence, quality being adherence to specifications and absence of defects; excellence being flexibility, customization, and the presence of nifty features and functionality.
Walmart’s response to Mr. Richardson constitutes a defect. The automated software selected the wrong boilerplate response instead of either finding one more suitable (unlikely given his inquiry) or passing it along to a live human being.
Imagine the designers had properly ranked their priorities, concluding that looking like an idiot might not be good for the Walmart brand. They might have (for example) included confidence scoring in their autoresponder software … scoring, that is, that calculates the likelihood the software had properly categorized the inquiry. With that in place, the system would have re-routed the question to a person, instead of returning a match that, while best, wasn’t good.
Pitfall 2: Missed supporting processes
Autoresponders digitize a business process (more precisely, they turn a manual practice into an automated process). They require content creation, content classification, message classification … straightforward stuff, at least until you have to do it.
It’s easy to miss the need for periodic content review, especially as it reduces the ROI. But without it, companies must rely on diligence to spot and fix out-of-date content. Requiring an owner and expiration date on every piece of content, with automatic notification for every message that’s passed its sell-by date? Easy, once you think of it.
Pitfall 3: Communicating from the sender’s perspective
Notice anything about Walmart’s message? Sure you did. Not one word is anything any recipient would care about. Even if it had been relevant to Mr. Richardson’s inquiry, it would still have been irritatingly long and pointlessly self-congratulatory. It’s written from the sender’s perspective, not the recipient’s.
How about you? When you institute a process change, are you clear about priorities? Do you and your team look for the missing ancillary processes? And your communications … do you write from the recipient’s point of view, or from your own?
Sure, you can learn from the many things Walmart does well, if you must.
But that requires humility. When I learn from someone’s mistakes I get to feel superior at the same time.
Wow Bob.
As usual, you have something interesting to think about.
This weeks piece while thoughtful, brought me first to laughter, then tears.
Re “Learning from an Easy Target”
See the article comparing IBM’s Jeopardy playing computer versus Wolfram Alpha. http://www.smartplanet.com/business/blog/smart-takes/wolfram-alpha-vs-ibms-watson-how-they-think/13912/
What you are suggesting Walmart should have done (and I agree), is NOT EASY to automate, and is not easy (nor cheap) to do with an all-human system. The best combination I have seen is from Interaction Corp., which uses “semi-automation” in its extremely sophisticated voice answering system that it sells to Fortune 5 companies. See http://www.interactions.net/cms/.
Rollie Cole
Pingback: Utility Bill Reduction – Stop Money Leaking Out Of Your Business | Blog Just 4 Klix
Looks like a good job for a cut down version of IBM’s Watson?
If you are in IT, Walmart can be fun to sneer at. But think of the one tech giant you would expect to get this right, and prepare to cringe.
Yes, I recent received a similarly non-relevant, excessively long, and annoying reply to a question I submitted to Google (the company, not the search engine).
I submitted a “contact us” form, one they specifically said I should use if I had questions about a notice they had just sent me. The answer I got was obviously a canned response to what their algorithm thought I was asking about – and was dead wrong.
Their failure was their hubris. The last field was a large text area where they asked me to describe the problem “in as much detail as possible” with encouragements to include error messages and notices. So I wrote the entire history of the issue, added pasted-in messages and notices from their web site and the email I got from them, and in general gave every bit of detail I thought was relevant.
This obviously confused their robo-responder. To make matters worse, after receiving the auto-response, I was left unsure if Google considered the matter closed or if a human would also respond.
After a sufficient amount of (wasted) time, I tried again, but this time described the problem in one sentence, using their language as to the nature of the problem and the action I wanted.
Sure enough, I was contacted by a human who took care of the problem right away.
Lesson to learn: don’t fall into the trap of putting blind faith in your systems ahead of adequate checks.
Whomever created that form and auto-responder system seemed to think their algorithms were so good, they could parse through a natural-language story mixed with technical jargon, and boil it down to the correct problem statement and edge use case. Well, they were wrong.
Now you have the opportunity to feel superior to Google as you learn from their mistake.
You’re welcome.
Good column, Bob – as usual.
I believe there is an additional dimension which could be added the dimensions of process optimization, namely, “political optimization”. This comes into play when the political forces, real and perceived, at work in an organization are such that all possible decisions or actions must be considered first in their political context. The result is that everything is at risk of being sub-optimized on the other six dimensions.