Cozmo App Overhaul
How might we help users discover all of Cozmo’s capabilities through the mobile app?
This is Cozmo. He can, among other things:
express emotions charmingly; he celebrates a win and pouts when he loses
autonomously path-plan and amuse himself
play games adversarially against human users
His “brains” live on his accompanying mobile app, and the primary target user is a child between the age of 7 and 12.
I was responsible for all UX design efforts on the Cozmo product line from June 2018 to company closure in May 2019.
The Cozmo team was tasked with overhauling the app. Users loved physical Cozmo’s personality but felt that “there’s not enough to do with Cozmo.” At the same time, most users did not regularly use most of the app’s features, leading us to believe that there was a discoverability problem.
Ultimately, the business and product goals of all redesign efforts was to improve customer satisfaction before and during the holiday season, which is when the bulk of sales happen.
Digging Into the Problem
Robot-side, because Cozmo is semi-autonomous, sometimes users had trouble understanding when they had control, when Cozmo had control, what Cozmo was doing at any given moment, and what he wanted them to do.
App-side, users weren’t accessing and engaging with a lot of content because they either didn’t understand how to unlock the content through the game mechanics, or weren’t compelled to.
The game design team had originally developed a pet-like system in which users had to feed, repair, and play with Cozmo everyday. Initially, this lifted engagement, but over time, the daily requirements became grating and was correlated with user churn.
In terms of repeated usage, open-ended features, such as Explorer Mode (shown above) or Cozmo Says, outperformed highly structured features such as Games and Tricks.
Design Challenge #1: Improve Discoverability
We had a few ideas for discoverability:
Improve onboarding so users grasp human-robot-Cube interaction, app navigation, and game mechanics
Reinforce onboarding concepts through design on individual pages
Restructure and redesign the home screen so that features are easily accessible in a carousel
Design Challenge #2: Improve Engagement
We tested out a few different solutions to the engagement problem:
Surface high-performing features on the top level
De-emphasize features associated with churn, such as Feed/Tune-up/Repair
Home Screen Explorations
Over the course of about 2 months’ worth of iterative testing iterations of the app redesign with 40+ kids, we found that:
Kids are very comfortable with scrolling and could find all of the content
Kids felt that there were “so many things to do with Cozmo”
Most kids had an “aha!” moment when they realized Cozmo’s Free Time showed them what Cozmo was seeing, and perceived Cozmo to be more high-tech
Final App Design: Before and After
We de-emphasized the meters
Kids were comfortable with carousel navigation
Kids are extremely literal when interpreting visuals, so we used very non-abstract titles and imagery
We separated Games and Tricks to reduce confusion
We locked Mini Games until users played one of the Main Games, where we provided video instructions on Cube placement, to reduce failure with Cubes
We added the “Play” meter to this view to reinforce that playing games fills the “Play” meter
Because users were often confused about how vigorously to shake a Cube, we made the hand-holding-Cube illustration in step 2 animate at about the speed that a user was supposed to shake the Cube
We placed a progress bar in step 2
Below is an app-walk through of the features that were changed in the app overhaul.
Project Impact & Learnings
After we shipped the app redesign, analytics from the 2018 holiday season product usage showed that:
Our Amazon product rating increased across all markets. Ratings in Germany, our weakest market, grew from 3.1 to 4.1 out of 5 stars.
During Black Friday week, we sold 40% more units than above forecast, so much so that marketing moved up sales targets. Overall, we sold about 50,000 units above the original forecast.
We also learned that:
While we created the version of Cozmo that earned the highest product review ratings in company history, week-over-week retention did not go up. We learned that our efforts created a much better first impression on users, which fortunately drove ratings and sales up, but that engagement is not necessarily the KPI to look at for sales. Thus, I would argue that overall, we succeeded in improving discoverability but not engagement.