When it’s Good to be Redundant in IVR Design…and Life

As early as I can remember learning anything, I was taught to strive for efficiency.  “Inefficient” and “redundant” were bad words, and employees who were laid off were called ‘redundant’.  We studied and admired Hemmingway’s succinct style in school, we learned to reduce our fractions to the simplest forms in math class, and in the age of increasing awareness of the earth’s environment we tried to use the least amount of paper, water, and electricity as possible.  Two-sided printing, half-size paper towels, hybrid cars with great gas mileage, speed dialing for phones, and automated transponders for highway tolls were developed and promoted to save us time and resources. 

Minimalism, Plain and Simple

Even as a young adult, my peers and I fought to be as efficient as we could at work and at home.  At my first post-college job, we competed with each other to digest our depositions into the smallest possible number of words.  Programmers who wrote the sparest code were called the most elegant.  When I met my husband, he was proud to own only four plates and a few forks, and he bragged that he could cook an entire meal with one pot and one big spoon.  I even had an apartment with a murphy bed that folded up, so I could live in one room.  Less was always more.

In the early days of IVR design – especially for speech applications – many designers, myself included, stayed true to this path and wrote and re-wrote our prompts to be as direct and simple as possible.  Fewer words translated to a simpler user experience that would be easier to navigate, we thought.  We were trained on example menu prompts that were short and sweet, like “chocolate, vanilla, or strawberry?” and “what size?”  Even our menus were as sparse as possible with no repeated options and nice clean branches.  We didn’t beat around the bush, we didn’t stop to explain, and a good voice talent could record a trim IVR application in an hour or two.

Like anything else, over time it’s inevitable:  we modify, refine, and mellow.  With two kids, a grownup social life, and a dishwasher, it wouldn’t be practical to have only four plates in our house – so we have a set of sixteen.  And although one of our cars gets great mileage, we also have a bigger car that can hold kids, dogs, and luggage and still make it up a hill.  Everyone I know has phone chargers upstairs, downstairs, in the car, and in the office – and many of my family members and friends have taken more of a zig-zag path to their careers than a straight line.  It’s easier to see now that there’s more than one way to get to the same place, and shorter is not always better.

Evolution of IVR Design: Purposeful Redundancy

In the case of IVRs and other automated systems, many designers also take a more modified approach to their user interfaces today.  For starters, early on we realized that recording multiple sets of numbers and letters with different intonation makes information readouts sound much more natural.  It’s inefficient, but it makes the IVR system sound better.  Sometimes the newer approach involves – gasp – a prompt that’s longer than we used to aim for.

We can sacrifice prompt brevity for clarity, explanation, and success.  This is a direct result of tuning and assessing many systems in production that were either experiencing technical issues or just weren’t being used – and then seeing the positive results we got after adding a few carefully chosen words or phrases in places where the users were getting stuck.  I’m not talking about the annoying legalese, warnings, and caveats we are force-fed on automated systems;  those are still a true waste of the user’s time.  But people will not object to a longer prompt if it’s clearer and helps them succeed when they’re making a transaction, getting crucial information, making appointments, and so on.

Designers even coach voice talents to slow down and repeat an address or phone number that the caller is likely trying to write down.  Research and usability studies have consistently shown that IVR callers want confirmations and the ability to hear repeats on crucial information, i.e. when there are significant consequences for incorrect recognition[i].

Although it hurts to hear a caller struggle to answer an IVR prompt you wrote and thought was brilliant, data doesn’t lie and real customers need to understand a question in order to answer it.  In the same way we need to expand our dishware or housing for a changing environment, we need to adapt our automated conversations if the originals aren’t working – no matter how aesthetically pleasing we thought they were.  When asking a consumer what type of loan they are interested in, for example, following the ice cream flavor model, you could try a succinct prompt such as “fixed, ARM, interest-only, buy-back, bridge, or reverse?” But it’s likely to cause hangups, swearing, muttering, silence, and confusion.  A live representative would never talk to a person like that; it’s just not enough information for the average consumer.

In the case of automated menus, it turns out that having options in more than one place can also help users because not everyone has the same mental model when they pick up the phone or go online.  Someone wanting to make a payment may go through a billing option looking for a mailing address, whereas another person with a question or complaint would be looking for the mailing address somewhere else.  During the tuning of a utility IVR application several years ago, it was discovered that only half of the callers wanting to make a payment arrangement selected ‘billing and payments’.  The other half expected to find payment arrangement in the ‘something else’ bucket.  So we put it in both places and automation went up.

Are shorter and more direct prompts automatically bad?  Should IVR menus have every option under every branch?  Of course not; I still like Hemmingway; I buy the paper towels with the half-sheet option, and I don’t like my time wasted on the phone.  But I also started parking farther away to get a spot in the shade, and I no longer count the number of words in my automated prompting.  I’m counting successful tasks completed instead.

Learn More about User Experience

[i] See AVIxD “When to Confirm” and “Repeating”.