It's Time to Accept Responsibility for Your IVR

Reposted with permission from the Interactions blog

by Dan Fox

There is a customer service revolution afoot. Consumers will no longer put up with poor customer service. As the number of product and service alternatives rises, the companies that provide the best customer experience will gain and retain, while the companies that don’t struggle to stay afloat. I swear I didn’t mean for both of those to rhyme.

Corporations are paying closer attention to the value of their customers. Those of you who follow Net Promoter Score are aware that detractors of your company (unhappy customers who feel trapped in a bad relationship with the company) are 3x more likely to speak about their experience and 2x more likely to speak negatively to 10 or more people. On the other end of the spectrum, Net Promoter Score leaders, i.e. those whose customer base contains primarily promoters (loyal enthusiasts who continue buying from the company and “promote” the company to people they know), are growing at more than 2x the rate of their competitors.

So how do you provide good customer service? Although that subject extends beyond the scope of this post, I can tell you where you can start, which coincidentally is probably the last place you’re looking: your automated system.
Most companies are focusing on adding channels to their customer support, such as social media, live chat, and smart-phone apps, and simultaneously adding features to enhance their contact centers, such as click-to-call or allowing premium status callers direct access to a live agent (as we’ve seen with the Chase Sapphire Preferred credit card). However, they all seem to be dancing around the subject of their IVR. Instead, they are looking for ways to boost every channel around the automated system in efforts to reduce the impact of a detrimental IVR or to reduce the call volumes into their contact centers. What’s the point of creating a brilliant multichannel customer support chain if your IVR is a broken link in that chain? I will now defer to the overused, yet appropriate cliché – putting a band aid on an open wound.
Why is my IVR so bad?

Hey, I’m glad you asked! Your IVR is so bad because it lacks the ability to understand your callers. Your callers are frustrated with your automated system before they even use it due to an abysmal track record and a lack of innovation in traditional Automated Speech Recognition (ASR) for the past 10 years. In fact, new research confirms that Natural Language Understanding is currently performing at a 65% accuracy on average. Callers expect:

  • The need to repeat information in the IVR that they give in the IVR
  • The need to repeat information to the live agent that they give in the IVR
  • Low expectation that they will be routed to the correct destination
  • Menu options that don’t fit their needs
  • Difficulty in reaching a live representative

Where does your ASR fall short? Do your customers have to repeat information to the live agent that was given in the IVR? Intelligent CTI will take care of that. Do they have difficulty reaching a live agent? If you truly care about your customers, you could design an IVR capable of addressing that problem. However, if your system suffers from speech recognition failures, an inability to granularly route, and inadequate menu options, you probably can’t do anything about that with the technology you’re currently using.
Why is my IVR’s automated speech recognition so crappy?

Don’t be so hard on yourself, it’s not your fault. Our traditional copper wire phone lines carry only audio quality that is easily understood by the human ear and brain. That “standard-definition” audio has a limited amount of data. Yes, even when someone is calling in on their iPhone 4S, with its “high-definition” microphone, their voice is still ending running on a “standard-definition” telephone system. It’s like the difference between hearing a song on the radio, and hearing it on those tinny 2001 ringtones – much more difficult to identify the song from the ringtone.
This is why, when you’re interacting with an IVR during your coffee break at Starbucks and you have a Scottish accent, you’re out-of-luck. Background noise and accents are an ASR’s worst enemy. Simple triggers can also set it off, such as the minute difference between b,c,d,e… (they all have an “eee” sound) when you are spelling out your first name, or words that sound similar, such as Boston and Austin.
The basic limitations of audio quality explain why, with traditional Automated Speech Recognition, you will not be able to capture e-mail addresses, log-ins, and other alphanumeric data to a high degree of accuracy. If your automated system requires any account identification or authorization, you have to find a creative way to dance around alpha-numerics. In fact, most IVR systems today focus on creative ways to make it easier for the automated speech system to recognize the caller, and less so on how to make it easier for the caller to use the IVR system. Even with data strings your ASR should be able to capture, it’s likely your callers will be forced to confirm or to repeat themselves. And if their responses include verbal disfluencies like uhm, ya know, or yeah, the likelihood is even higher. For most callers, this is a major frustration point.
While it’s true that Automated Speech Recognition and Natural Language are getting better every year, there will always be a disparity between ASR technology and human understanding that prevents your callers from communicating in their own words. Your caller doesn’t want to Technical Support -> Home -> Internet Router, she wants to fix her broken Internet! Natural Language, at its best, still doesn’t allow your callers to speak in their own words. Directed Dialogue forces your callers into speaking in a way that works for the system. You’ve heard them before: “How can I help you? You can speak to me in full sentences, but you can only say ‘make a reservation’ or ‘cancel an appointment.’” Clearly, I’m exaggerating, but my point remains. Your callers want to express their problems in their own way because they believe that their call is unique (which is why they are calling – they probably already looked to complete their request on your website) and that your automated system won’t be able to meet their needs.
What am I supposed to do?

Call Interactions. No, I’m just joking. But seriously. You need to think creatively. ASR works in many areas, but as I explained, there are many areas in which it fails. Supplementing your ASR with human-assisted understanding provided by several thought-leading companies, such as Interactions, can make your IVR work. (If you are unfamiliar with human-assisted understanding, check out a demo.) This approach will eliminate the pain points, providing greater understanding that will enable you to build more functionality into your system. That being said, many IVR companies have tried to build human-assisted understanding IVRs themselves, but have been unable to meet their ROI requirements. When they quantify the benefits to the call center infrastructure and compare that number to the price tag for improving their system, the math doesn’t work. But at the risk of sounding like a marketing professional (which I am), I have to tell you that Interactions has cracked the code. Our many customer-centric clients can attest to that.
Do I need analytics to tell me what I’m doing wrong?

Yes, of course you do. You need to know the route your customers are traversing. Are you receiving a heavy call volume for something that could easily be accomplished in self-service? On that note, you also want to know why they are hitting the contact center. Is the self-service functionality not prevalent enough? Is there even a self-service functionality on the website for that command? Perhaps an abnormally heavy call volume for that command can even alert you to a bug on that part of the website.
Analytics allow you to have greater control over the flow of your customer support channels, identify bottlenecks, and better design your support structure to optimize customer experience and improve your economics.
You’re not telling everything, are you…

You’re right. Analytics are great in theory but most companies and self-service analysis providers treat customers like sheep. The predominant thinking is “I have this huge call volume and these expensive agents, I need to save some money, so let’s build this menu option here, and this self-service functionality there, and link them to the website here. In turn, we’ll increase containment, cut down Average Handle Time (AHT), and increase customer experience.”
That’s all well and good. However, I return to my original point:
You can gather and analyze all the customer data you want, but if you are still relying solely on ASR to respond to all your customers’ needs, you are missing a big piece of the puzzle. Instead of forcing the customer to communicate in unnatural ways that suit your system, you need to build a system that accepts and responds to what they have to say. You have to accept responsibility for your communication mediums and stop forcing your customers to speak in the way that’s best for the IVR. It’s time to get creative. Stop thinking about your IVR as a gateway to your representatives, and start viewing it as a critical, strategic initiative.

Recent studies have shown that customers are willing to interact in self-service if the automated system works for them. Unfortunately, traditional speech-based IVRs cannot meet customers needs and perform at a 65% accuracy on average. In a new whitepaper entitled “Broadening the Conversation” Interactions’ CTO, Yoryos Yeracaris takes a look into new Adaptive-Understanding™ technology that transforms self-service into a value-added service in customer care.