Remember the bad old days of customer self-service? Automated systems would present you with five or more different options — any one of which could lead to five or more other options. With touchtone systems, accidentally pressing the wrong number at a prompt could result in instant death by disconnection. Then came the speech-recognition systems, with their own specific tortures: creepily cheery voices reciting five or more options; systems that went into lockdown if you sneezed into the phone. Let’s not even talk about systems that couldn’t understand your accent, whether you were from another country or just another region.
That was then, though, right? Wrong.
Interactive voice response technology is the foundation of self-service technology, and it’s been around in various forms and formats for a good two decades. The technology is mature and stable — a condition that suggests it should deliver a glitch-free experience for users.
Unfortunately, that is still not the case. For one thing, innovations in speech recognition and IVR systems tend to creep along. One change, though, is taking hold. Namely, there are better tools for detecting problems with an IVR menu or speech-recognition system, so companies can fix them sooner.
What’s Old Is New
One wouldn’t think such tools and processes would be necessary, but many of the same weaknesses that bedeviled the industry 10 or 15 years ago still exist, according to Ken Landoline, principal with Synergy Research Group.
The more egregious glitches have been remedied.
“I remember systems that would disconnect women that had high-pitched voices,” Landoline told CRM Buyer. “The systems would interpret their responses as a certain tone and automatically hang up.”
The algorithms have improved enough to avoid that problem, he said.
However, “they still have quirks,” said Mark Camack, principal with IVR Doctors, “and even some of the very best systems don’t always work, even when a consumer says a word very clearly — there are still hiccups.”
Such issues tend to loom large with callers who don’t like IVRs in the first place.
“They are already going into the transaction with the idea that they will be inconvenienced or won’t get what they want,” Camack told CRM Buyer, “so the minute something happens — the system asks the caller to repeat his answer, for instance — the caller thinks ‘Oh, here we go again,’ and automatically zeros his way out of the system.”
The end result is that the company has spent a lot of money on a system that its customers still won’t use.
Poor integration is also still a significant problem. Unbelievably, it is still common to see misapplied and poorly integrated CTI technology, said Leah Eyler, a business analyst for customer interactive solutions at Dimension Data.
Any standard CTI package that is part of an ADC system will work well when properly implemented and integrated, Eyler told CRM Buyer.
Being asked more than once for account information in a call is a common sore point, for example. Better integration with a back office customer database is an easy solution, Eyler said.
“The first order of business, I think, for many companies, is to focus on getting their existing technology to work correctly — and then focus on new developments in this space,” she said.
“The last quantum leap for IVRs was the shift from touchtone to speech interface,” Landoline noted.
Since then, improvements have been occurring at a snail’s pace, but the industry is not standing still.
Now, enterprises are eager to accommodate mobile phone users, who tend to be younger. Younger users are less likely to be irate when confronted by an IVR customer service option, Synergy’s Landoline said — and less likely to automatically zero out looking for a human.
“So, companies are pressing vendors to deliver speech recognition systems that can handle the mobile customer — someone who is walking down the street or driving a car while conducting a transaction.”
Background noise is the big challenge in this scenario.
At the moment, applications or tools that help organizations identify what is not working with their systems are making the biggest impact on the market, said Donna Fluss, president of DMG Consulting and author of The Real-Time Contact Center.
“These are optimization applications from IVR vendors and standalone providers,” she told the E-Commerce Times. The category is loosely defined as “customer experience analytics,” or CEA –“an emerging application area which looks at the customer self-service experience.”
Usability, or human factors testing, is now an integral part of the development process, agreed IVR Doctor’s Camack.
This practice has been around for decades. However, it has become “chic” — to use Camack’s word — among enterprises now, and it’s sought out in the script writing and system design phase, not after the fact.
“It has gone from an afterthought to something that is now considered essential,” Camack said.
These tests vary depending on the task that is being automated, the user base, and the supporting infrastructure, noted Dimension Data’s Eyler. “In general, though, you just walk through a transaction. The speech app will pose a question — the caller gives an answer. Do the two match up? How long does it take to respond to the question? Are there other possible answers that are not addressed by the menu of choices?”
Sometimes the results can be surprising even to experts in the field.
For example, one project Dimension Data conducted for a client called for adding a new speech module to an existing IVR System, Eyler recalled. The task the company was trying to automate for users was a complex one requiring an open-ended dialogue between the system and the caller.
Usability testing uncovered “clunkiness” in the IVR System — and a flawless implementation and integration of the speech app. In short, the menu of options available was not logical to the task — but once users got through the meandering queue they found the system easy to operate.
The company wound up making changes to the older, established IVR technology, she said, but it left the cutting-edge — and supposedly still buggy — speech application alone.
Great article, Erika!
It’s true that speech rec is slow to improve. But I’m convinced that that’s not a flaw in the industry or any reluctance towards innovation; I believe it’s because speech rec is dealing with human beings.
Sure, we can and should tweak the systems; IVR improvement should be an ongoing process, not a one-time implementation. The bottom line is, though, that humans will always trip up the system. We will sneeze, yell at our kids, start giving the wrong phone number, and say "um" a lot. (Ever had an "um" trip up a system?)
So what do you think the next step is? Can we ever perfect the system, or can we simply train our users to be more exact on their end? Will we ever hear the end of "I *hate* those automated speech systems! Just give me a human being!"
I wonder if the answer isn’t to put a little more human in the system. What do you think?