I designed a voice-based conversational phone system for customers who call the Home Depot stores and contact centers.
From 2017 to early 2020, I worked as a user experience designer for the contact center experience team (CCX). The CCX was tasked with designing, developing, and maintaining all systems used by the various contact centers within Home Depot. My team, specifically within the CCX, was asked to solve the following problem - “As a customer, the choices in the IVR are too confusing and numerous to get me to the right place.”
This is one of three different journey maps I constructed to highlight customer pain points when they called into our stores and contact centers. This specific scenario is focused on reaching the online contact center.
Home Depot customers don't want to listen to long and confusing menu prompts that eventually lead them to the wrong person or place. They want to be directed to the right associate, the first time, and get their questions answered quickly.
This redesign allows customers to tell us what their inquiries are instead of asking them to guess from a series of ambiguous menu options. Through this process, I gained a ton of experience with conversational design principles and the voice design process at a company where conversational design did not exist. How do you go about testing for voice? What are the steps necessary to create a great conversational experience? Who were the industry leads and what lessons were learned?
But we know that conversational design is KEY to creating a great voice experience.
In designing the new system, I utilized a great deal of conversational design principles to make the interaction between Home Depot customers and our system as painless and helpful as possible. Some of the ways I was able to this was…
Leveraging the customer’s information to personalize the call as much as possible.
Being transparent with the customer on what capabilities are available in our system.
Handling errors as gracefully as possible and allowing customers to escape to a person at all times.
Adding confirmations when necessary so customers feel confident in the system.
There were many other findings from our research and testing, but we also worked with our data science team to further improve the system by…
Training the system to accurately pick up customer’s inputs and correctly match intents.
Iterative testing and feedback loop to continuously improve the customer’s experience in our system.
In this end-to-end process, I gained a ton of other skills that ultimately helped me become more than just a designer.
Build v. Buy Process
I convinced my product team to conduct a build verse buy analysis after our initial problem discovery. We knew what our ideal solution would look like, but we needed to decide whether we were going to work with a third party vendor to implement the solution or build the solution in-house. This was a few months of evaluating internally whether our engineering team was capable of implementing the proposed solution. We spoke with multiple vendors and ultimately evaluated the pros and cons.
**please note that numbers and information have been changed to ensure confidentiality with The Home Depot.
2. Grammar Studios (Information Architecture exercise)
Because of the lack of data we had when we first started implementing our solution, we had to be scrappy and figure out a way to get data quickly. We needed insight into how customers would react to certain phrases and words. It was a way we could initially train our intent model. We had around 5 to 10 participants in a room and would ask each person to list on sticky notes the first thoughts or phrases that came to mind when a certain word or phrase was given. For example, if the world “tool rental” was given, a participant might write words such as “rental equipment,” or “borrow a tool.” These were a huge success and ultimately the activity became more than just a way to gather grammars for an intent model. Designers across the company leveraged and customized this activity for multiple mediums outside of voice design. They were using this activity as a means to gather quick and dirty data on the initial reaction to anything whether it was a photo, an interaction, a word, or even a sound.
3. Design Sprints
I facilitated two large design sprints with our stakeholders for both the order status self-service functionality as well as the pricing and availability self-service. We modified the design sprint to fit our own specific needs and the needs of our stakeholders. Our sprints consisted of identifying the right problem, understanding the problem, solution/ideation, and storyboarding. The last days were centered around testing and design iterations. These sprints did a phenomenal job getting stakeholder participation and alignment on what our ideal design might be.
OUR PROCESS
The balanced team initially followed the Discovery and Framing effort but had to adapt and used the design sprint method for our continuous improvements. We are currently at 600+ stores and have plans to continue rolling out to the remaining stores AND the contact centers this year.
THE OUTCOME
In December of 2018, we went live with an entirely new voice-based IVR solution that:
average time - decreased average customer time in IVR from 180 seconds to 18 seconds.
reduced greeting - shortened the initial greeting from 1 min and 40 seconds to 7 seconds.
reporting visibility - went from zero reporting to creating a reporting console.
self-service - allowed customers to check their order status and find pricing/availability information.
customer satisfaction - from 3.2 to over 4.5 out of 5.
WHAT DOES IT SOUND LIKE TODAY?
Here’s an example of what the product availability self-service sounds like. Please keep in mind that this is our most minimum viable product and does not reflect our ideal design.