Problem Statement

We’ve developed a MVP that lets users upload text documents, preview and edit their infographics.

How do we go from here, to a Beta experience users need and love?

Understand & Analyze

Understanding how, what, and when our early users  use Thoth for.

For Alpha, I set up various qualitative and quantitative feedback channels to track users' satisfaction and feedback. This included:

In-depth interviews

Quick and dirty interviews

NPS Surveys

Reviews/Feedback/Report a Bug

Throughout Alpha, I personally followed up with everyone who joined via e-mail. I spoke to 15 individual users, and managed several Thoth client demos/projects/events.

The goal was to understand more about our users’ needs in their workplace, what makes our users tick, and dealbreakers when it comes to using the current platform.

While the overall experience was relatively straightforward for our early users, there were some common frustrations and wishlisted features brought up:

Quality of automated summarizations were disappointing

More than half our users said that they were disappointed by the quality of the summaries, rating it a 2-3 out of 5.

Immediate feedback to edits is a must

During user testing, people often tried to click on the infographic to edit it, and were confused when there was no feedback. They were also frustrated with the waiting time of 3-4 seconds every single time they clicked on 'Save Changes' to see the preview.

Layouts feel repetitive, with little formats/aspect ratios to choose from

Layouts started to feel boring to users after 4-6 documents, and they wanted more variety. There was also requests from a few users to provide square & vertical formats, as well as multi-slide infographics for LinkedIn.

Only PDFs? We want more input formats!

We only accept PDFs at the moment, but we noticed that many people prepared their documents in word documents. Some users also asked for a transcription service to translate audio immediately to infographics.

Some users have no documents on hand to try Thoth

"I didn't know what to test, but I just took a random XXX" This was a frequent occurrence when I was speaking to the users.

Corporate customers wants more brand customization and differentiation.

Corporate clients wanted the infographics to match their brand guide - which meant custom fonts, a specific illustration/visual style. They were also concerned whether their infographics will look the same as other clients after a while.

Features Prioritization

As a group brainstorming workshop, the team brainstormed on how to address these issues.

The end result was a (long) list of features, which we then prioritized into different buckets based on their impact and effort.

Brainstorm & Design

Feature Highlights

An WYSIWYG editor for immediate feedback

Changes are now almost instantaneous. Users no longer have to wait 3-4 seconds each time they saved changes for it to be reflected in the preview (which was done almost 20-30 times in a single setting on average).

Additional input formats for word doc, audio and video formats.

More file formats and integrated transcription processes mean users spend less time on formatting their document, to focus on key summaries and building their infographic.

User curation for more precise AI-driven summaries

How do we design a software that’s highly dependent on the performance of AI - and build for its success?

A part of it is managing user expectations, which we are improving on through the copy and onboarding process; the other part is letting users guide the AI.

We included a text editor for users to edit their text based on recommendations (especially for audio files), and keyword tags to let the AI know specific areas to look out for.

Demo articles and onboarding tutorial for first-time users.

Users can learn how Thoth works via an interactive walkthrough. We also included demo articles, where users can test out Thoth with limited features.

Growing our user community through user-led referrals

Part of our growth strategy is building a user community as we scale. This allows us to connect with our early users, get more user insights and help us interate on our product further.

We introduced a referral scheme that allows us to scale our community sustainably.

Key Learnings

Designing alongside the state of AI

How do we manage users’ expectations of AI? How do we make their experience better, despite AI’s hit-and-miss performance? How do you do user testing for AI without actually building it? These are constant challenges that we face, improve and learn from.

Building design systems for scale

I worked closely with our frontend engineer to design and build up our design system and library of reusable components, making it more efficient and consistent both in design and in code.

Adopting a pragmatic approach to design and research for rapid iteration

When faced with limited resources and a tight timeline, sometimes it is more efficient - and cheaper - to simply push features out and iterate as we go along. As a compromise to in-depth user testing, we set up analytics and systems in place to track key metrics such as time spent on editing and NPS scores. Whilst not able to push for an ideal research process for all features, The goal is always putting users first - and I am setting up and advocating for stronger research and analytics focus in our design-development process.