American English pronunciation app


SpeechAce is an American English pronunciation app that helps ESL speakers improve their accents, but for many users, general usability problems and a mismatch of features to user needs made attracting and retaining users difficult.


Construct a persona based off user interviews & feedback, define user stories, evaluate the app for usability issues, re-design problematic interactions, and work with the product manager to build a product roadmap:

  • Persona
  • Expert review
  • Re-design
  • Product backlog


SpeechAce received an update to existing functionality, and a short list of newly identified features that lined up with core user needs were added to the product roadmap.


Usability was increased substantially by focusing on the persona I constructed from interviews with existing users. The SpeechAce team has implemented some of the proposed changes and built a roadmap off of the features backlog.


Based on the feedback from an initial round of user testing that had already been conducted, users were clearly having a hard time using the app to achieve their goals.

Next steps were constructed around solving for:

  1. Who is the user? Develop a rough persona to better define user motivations.
  2. What can users get out of using SpeechAce? Compare current features against user needs.
  3. How can users accomplish their goal (better pronunciation)? Prioritize new features and improve usability with a lightweight re-design.

The widespread confusion about how pronunciation was being graded told us that revisiting the pronunciation feedback graph in particular would be a good idea.


Previous interviews of SpeechAce users helped us narrow down a rough sketch of who the typical SpeechAce user is.


With the new persona as our bullseye, I wrote a handful of user stories to better understand what SpeechAce users want to get out of their experience using the app, and used that process to help define a couple of additional features.

I sorted all tasks into a red routes chart showing frequency vs ratio of users, helping to visualize feature priorities. Focusing on the features in the red boxes will provide valuable functionality that helps users accomplish their goals.


The feedback graph on its own was particularly frustrating for most users. Dropping the individual phoneme grades in favor of evaluating whole syllables improves the speed of comprehension, as well as using a more universally understood scoring system with secondary visual color and emoji indicators to reduce cognitive load.


I also outlined a short list of high-priority features to follow up with that would meet core user's needs:

  • Alternative email address signup & login methods
  • Whole sentence practice and grading for fluid speaking
  • Track and test the user over time (spaced repetition algorithm?)
  • User-created word lists for customization

Suggestions for improving specific pronunciation errors is an NLP (Natural Language Processing) problem which requires a more sophisticated development solution than SpeechAce can support currently, but having identified the need, the SpeechAce team added that feature to their long-term product roadmap.