Monday, March 14, 2005

Say What?

(Warning, long and potentially boring post below. Enter at your own risk.)

A couple weeks ago, I had some fun writing about my experience with Alex, the voice of the USAirways baggage inquiry system. (I finally received by suitcase sometime between 10 and 12 on that Sunday; it was sitting outside my door when I got back from running some errands.) But my problems with Alex weren't unique to that particular system, and I'm certain I'm not the only one who gets frustrated by such poorly designed systems.

If I get more frustrated than the average consumer, it's because I've been programming interactive voice response (IVR) systems for over ten years. I think I have a handle on what makes one system better than another. And while I've only been working with speech recognition for a little over a year, I believe my experience carries over. I'm pretty good at this kind of programming, at least when using the development system that I've got the most experience with. However, my job is not something I like to tell people about, because almost invariably it goes like this:

Other Person: So what do you do?
Me: I'm a computer programmer.
OP: Oh, what kind of programming?
Me: I work on voice response systems.
OP: What's that?
Me: Well, you know how when you call your bank or credit card company and you have to punch in your account number and all th . . .
OP: [interrupting] Oh, I HATE those things.

I've been hearing that for ten years. It's depressing sometimes.

Two pieces of information from a conference call that I just sat through. 1) the users of the new IVR system I just implemented reported that they are happy that the system in question was touch-tone based (or, as we say in the business, "DTMF only"); they really dislike the speech recognition systems other companies have implemented. And 2) while discussing how to implement a change to allow more users to get further into the system without being forced to an operator, one of the business contacts asked "well, what's easier [to program]?"

So, first of all, why is it that the users hate speech reco so much? I can only guess in this case, but maybe it's because the systems and dialogues were specified (I almost said "designed" but I don't know if that much thought went into it) by people who were not thinking of the users so much as they were thinking of themselves. Too much in-house jargon gets used. Too many assumptions are made about how much the caller does or doesn't know. Too much imitation of what is already out there. I've been in too many planning sessions where people come up with the menus and prompts that sound just like what they heard when they called Company X. Did you like company X's system? Probably not, but there seems to be this mindset that every IVR system should sound like the last one you heard.

Speech Reco looks easy on the surface, and it does provide some capabilities that DTMF does not -- input of letters, of course, or selection from a list that is just too big to read off along with the requisite "press 1 for ..."; e.g. city names, as used within an airline system. (of course my pal Alex didn't seem to do so well with letters or cities.) What is hard about SR, and what usually gets short shrift, is design. Given any specific piece of information being requested from the caller, you have to anticipate any possible response. Even when you are expecting the caller to say "yes" or "no," just on the "yes" side you have to try to allow for: yeah; um, sure; yup; yes ma'am, etc. I don't know that I'd go so far as to program "true dat" (as was suggested to me recently), but maybe "you bet" ought to be in there.

The questions you ask have to have clear options as well. Quick, what would you reply to the following:

Would you like benefits information for: In-Patient Hospital, Outpatient Surgery, Out-Patient Hospital non-surgical, Emergency Room, or Behavioral Health?

Or: Are you a member currently enrolled, a perspective customer who would like more information about this plan or a healthcare provider?

Aside from not knowing what a perspective customer is, I'm saying "operator" to both those prompts. I'm probably saying something else that might be a little rude too, hoping that they might be using a call monitoring system.

Now about the "what's easier" question . . . Arrgghh. During that conference call, I politely mentioned that it shouldn't be a matter of what is easiest, and that in fact all the options being discussed were equally easy, but that it should be a matter of what is best for the callers. This sort of thing happens way to often, I've found -- the concern is on what is easy and/or quick to implement, or worse, what will keep callers from reaching an agent. Yes, nothing says "customer service" like avoiding your customers.

When I first was offered the chance to work on IVR, I was of course well aware of the general public's opinion. I (naively, perhaps) though that I had the chance to make a difference, to program something that people wouldn't hate so much. Unfortunately, corporate America isn't structured that way. No, I'm just the programmer. Other people get to decide what the systems sound like, and how many menus there will be, when it might be OK to transfer a call, etc. It's call center managers who only want to keep the number of incoming calls to a minimum -- zero being the ultimate goal, it seems. Or other management types who've been told how much better speech recognition is, how it will take care of all those people who only have rotary dial phones so they can't interact with DTMF based systems. Yeah, that's the reason people aren't using your IVR systems, they all have rotary phones.

Still, I idealistically soldier on, because I actually do believe that this technology can be used for good, not evil. I know about the cost differential between a call handled by an automated system and one handled by a human. There are many times when I, and I believe others, actually prefer dealing only with an automated system. If the information I want is easy to get to and everything I need to know is provided for me, then sure, I'm all for it. The technology is the easy part, actually. The difficult part is designing the system so that it is useful and useable, for both the customers and the business. I think it takes a combination of marketing, customer service, psychology and technical knowledge to do it well.

In a previous employment situation, I fought hard for the creation of, and finally became part of a committee that reviewed all of the scripting for an IVR system, rewording things as necessary, adding consistency to menu options (e.g. "end this call" was always 9. "return to the main menu" was always 8). We had representatives from marketing and customer service and IT. We studied the usage patterns of the existing system to find out what information people were requesting most often, and rewrote the system to make it easy to get that information with a minimum of key presses. By increasing usage of the system and decreasing average call length, we saved the company, on paper at least, several million dollars per year. I know it can be done right. I just don't know why corporations won't invest the time to do so.

You can sometimes see evidence that they're moving in the right direction. Somebody probably told USAirways that a male voice was more appropriate for their baggage inquiry system for some reason. I'm sure I could locate studies that indicate which gender is preferred for various types of systems. So they got that part right. Why thny did they get everything else so horribly wrong? Why do they start the call with a female voice, and DTMF only? Why do I need to be "introduced" to the disembodied voice on the the other end of the line? If Alex doesn't understand me when I say "PHF," why doesn't he just say so? Instead, he just jumps ahead to "That's OK, I can look it up another way." I could be wrong, but it seemed like I was the one being blamed for the system not understanding me. In fact, as I recall my several conversations with Alex, I don't think he ever said that he didn't understand, he just kept going with what *he* wanted to say. In that way, I guess he was more lifelike than I'd thought; he wasn't listening, he was just waiting to speak.

Perhaps USAirways has valid reasons to keep me from speaking to a real live person, but to completely ignore the fact that somebody would say "operator" or "transfer" or anything similar is just ridiculous. Acknowledge what I've said, tell me that the information that I've been given is the most current information available, something. Anything. Please?

To make matters worse, on each call, Alex encouraged me to use the web, implying that it's so much better. Since I was in a location with no access to the internets when I first called, that wasn't an option (hmm, they believe that millions of people have rotary phones, but also that everybody has an internet connection?), but when I decided to go to the local coffee shop and use the available wi-fi connection, I took the slip of paper that had my claim information on it. I brought up the USAirways website and had to guess where to find the right page for baggage inquiries. Once I found it, I was presented with a form to fill out - name, address, email, claim number. And then another "required" field: flight number. OK, flight number wasn't on the slip of paper I had, Alex had never asked me for the flight number, but all of a sudden, the allegedly better website needs to know? Luckily, I was able to pull up the the confirmation email from Expedia in another browser tab, and I copied and pasted the flight number into the form. Then I clicked on the submit button and all I got was a "thanks for sending your information." That's it? They drive me to the web to use a non-interactive system? Happy Bankruptcy, USAirways. How's that working out for you? At least we know the US in your name doesn't stand for "usability." How much would you like to bet that the web design team and the phone system people never talk to each other?

Anyway, if you meet me in a bar somewhere or something, and ask me what I do, don't hate me when I tell you, OK? I'd love to join some forward thinking corporation and help implement my ideas, but in the meantime, I'm only doing what I'm told. Of course, if you want to offer me a job, that would be cool, too.


Post a Comment

Links to this post:

Create a Link

<< Home