A major Telco engaged us to help understand the opportunities for voice interfaces across multiple product verticals, including mobile billing, customer support, handsfree calling, television control, and account management. By identifying opportunities with high product impact potential were were to help them develop a strategic product roadmap.
This 6-week long discovery project began with stakeholder interviews, helping us to ground our user research activities in the products and services we were exploring.
To help us learn, we engaged with customers across five user segments and two Canadian cities. The product and service offerings were different in eastern and western Canada so tests were done in both Toronto and Vancouver.
We spend an hour with each participant, starting with a context interview, and moving into validation testing with a WoZ prototype of our Telcom assistant. This was followed by a debrief interview and card sorting exercise.
To help us with validating some of our early hypotheses we used our internally developed CUI design and prototyping tool Simili to design and test two different prototypes that each focused on different domains.
You can read about SImili's development right here. During this project we were simultaneously developing Simli, as we used it to prototype we identified high value features and then validated them for ourselves.
Simili allowed us to design and test various conversation flows without having to build and train a robust NLU model, allowing us to fail quickly and iterate on our learnings.
Most desired functionality
When we returned to Toronto we spent a week synthesizing the data from our interview notes, test transcripts and card sorting outputs and summarized them in a synthesis report. Of the 6 domains we were initially asked to explore, that were two that were clearly the areas of opportunity, with a third (television) identified that was not initially anticipated.
Television & streaming
The TV guide is frustrating and a conversational assistant that can help manage recordings and navigate programming options was well received by our testers.
- Conversational channel surfing
- Setting up recordings
- Setting reminders for programs /sports
- Browsing movies and television shows
- Recommendations based on history
- Searching for titles, genres and actors
Another point of customer frustration and something our testers said they would like to hand-off to an assistant. This was seen as a better alternative to calling unresponsive IVR systems.
- Checking usage for data, SMS, minutes etc.
- Checking overage charges & dissecting bill
- Plan details / Plan alternatives
- Adding data / Easy Roam
- Potential for step-up recommendations if appropriate
An obvious feature for a smart speaker with significant benefits to the customer. Taking advantage of existing Telcom products and services.
- Calling making / receiving
- SMS send and receive
- Voicemail - play / erase
- Caller I.D.
- Connect and talk to technical and customer support with call back option
- Find my phone
A conversational guide
During our Vancouver validation testing it was quickly apparent that controlling the television experience was a highly desirable domain for the Telcom assistant. This was exciting because it wasn't on the radar of our stakeholders.
TV guides are a perennial frustration that have never been truly solved. The more complicated the menus become, the more frustrating they are to interact with. The customers we spoke with suggested that talking with their cable box and controlling it through a home hub was something they would expect in an assistant.
Because Simili is a really rapid tool to design and iterate with I quickly designed some sample dialogues of television control and search so I could validate desirability with the rest of our scheduled testing sessions. Below are some quotes from our participants about their favourite feature:
What is your most desired feature?
When we asked this question in the closing interviews of our Vancouver sessions it was extremely evident that the television and streaming domain was highly desirable for customers.
"If it could search on television and serve up things that I really like, if it's going in that direction, then it would be nice. It becomes sort of a companion" -Charlenne
"Just saying it for TV controls is really cool, for turning off. Scheduling would be great." -Danielle
"The TV Assistant, I'd tell it to record my shows, find out when the next show on tv is, find the news, sports highlights, and the number one song in the top forty charts." -David
We collated our users' utterances for further analysis, the sheer amount of television-related requests was telling. Utterance data like this can be used to identify missing intents or unanticipated use cases. It can also be valuable training data for our interaction model.
Our research synthesis and summary conclusions included insights into the desirability and usability of voice interfaces and highlighted several validated product directions that would be good fits for voice-based technology, including the discovery of an entirely new opportunity in the television domain that showed significant promise in our user testing.
Since this engagement, we’ve learned that this project directly informed our client's product roadmap and has led to the development of several successful in-market products.