A sign up flow that resulted in a 130% increase in conversion.
Project overview
Korbit is an EdTech start up in Montreal that aims to offer high quality, interactive and personalized education through Korbi, an AI tutor. As part of the platform redesign, we tore apart their original sign up and curriculum personalization flow and came up with a better version that resulted in a 130% increase in conversion.
Team
My Role:
Solo UX designer
In collaboration with:
CEO, CMO, COO, Data Scientist, Front-end and Back-end developers
Problem
1. Beginners in data science bear the burden of personalizing their own curriculum
2. Background questions don't serve a purpose
3. Poor Usability (too many choices, unclear button states, not accessible)
Goal
Allow both beginners and non-beginners to create their personalized curriculum in our sign up flow easily. Improve the user retention and engagement rate.
Solution
1. Segment beginners from non-beginners
2. Deliberate choice of questions that serve a purpose in personalization
3. Revamp UI
Impact
130%+ in conversion
The Problem of Our Original Flow
- Students bear the burden of personalizing their own curriculum - they were asked to select data science skills base on the real world problems they want to solve, but users (especially beginners) have a hard time making such decisions.
- Poor Usability - the original flow was developed without a designer so there are issues like unclear button states.
Detailed analysis under 'The Process' below.
The Problem
This was their original sign up flow:
.png)
The general feedback when it was shown to focus groups were good, and it also performs better in terms of user retention in an A/B test with a frictionless sign up flow (create an account only). However, no user testings were done when it was first designed so the usability issues become apparent as more and more users go through it.
After gathering both quantitative and qualitative feedback, we can conclude that:
- A lot of users (especially beginners in data science) don't know what skills to choose - from our data, a large amount of users choose 1 or all of the problem and all of the skills offered to them
- The UI is confusing - the clicked state of the Yes/No button is not apparent enough
The Goal
Allow both beginners and non-beginners to create their personalized curriculum in our sign up flow easily. From a business perspective, we hope to improve the user retention and engagement rate.
01
Discover
Research Methods
To discover the reason why our sign up flow is underperforming, we started with research:
Qualitative
Interviews with existing users
Quantitative
Clicks, Session Recordings
.svg)
Market Research
Friction-based sign up best practices & how other companies do it
Findings
Beginners in data science don’t know what skills to choose for their curriculum
.svg)
System behaviour not as expected - 1 user chose ‘not familiar with maths’ and got a math exercise as the first question
.svg)
Difficulty making decisions due to the large amount of choices -
a lot of users chose 1 or all of the presented options
Quick heuristic evaluation of our old flow reveals several usability issues.
02
Define
Defining the Problem
After analyzing our research, we then decided on the 3 top problems to focus on for this iteration:
Beginners in data science bear the burden of personalizing their own curriculum
Background questions don't serve a purpose
Poor usability (too many choices, unclear button states, not accessible)
Defining the Goal
Based on the problem outlined above and our target users (professionals with different background and learning objectives who want to upskill in data science), we defined our goal as:
Allow both beginners and non-beginners to create their personalized curriculum in our sign up flow easily. Improve the user retention and engagement rate.
03
Develop
Brainstorming with HMWs & Converging
We spent some time as a team brainstorming ideas with HMWs, and then, went over each of our idea, discussed whether or not to keep it base on its impact & effort, and merged similar ones together. We arrived at the following decisions:
Segment students into beginners vs non-beginners
Students looking for specific AI skills will be given the power to customize their curriculum
Ask meaningful questions
All questions should serve a purpose and actually personalizes their curriculum
Let Korbi (our AI tutor) guide the entire process
For a seamless integration between signing up and learning with Korbi
I like to write down thoughts and reasonings next to the screens.
03
Develop
Wireframes and Feedback Round #1
To better visualize how our ideas can be translated into a flow, I sketched out 2 flows that incorporated most of the ideas we agreed on, and wrote down all the questions needed to be addressed in the next meeting. We met again, decided on which direction to take, and discussed the flow in more detail, including:
What questions to include and their purposes
(how they will affect the curriculum)
Ordering of questions
How to segment beginners from non-beginners
Due to NDA I've omitted some details here
High Fidelity and Feedback Round #2
Once we've agreed on the high level flow, it's time to start working on the UI. I designed the high-fid version on Figma and met with the team again for feedback where we went into details like wordings, iconography and UI elements.
04
Testing
5 Internal + 5 External Testings
To meet the launch deadline, I first tested with 5 of my colleagues, and later on tested with 5 external users again. Here are a summary of findings:
It's a significant improvement from the original version
Additional questions gave rise to a more personalized feeling
05
Experiments
A/B/C/D/E Testing
Being a data driven company, we carried out A/B tests with 5 variations of the flow (aka A/B/C/D/E tests). We wanted to know in terms of conversion:
Is the new design is an improvement from the original one?
Will additional (potentially optional) questions hurt conversion?
To our surprise, we concluded 3 weeks later that:
The new design brings a 130% increase in conversion, the highest among all the variants
Shortening the process by removing additional optional questions actually hurts conversion
05
Experiments
Next Steps and Reflections
Continuous Iteration
A design is never done. We're still iterating on it, and to date we've tweaked icons + copies in order to continuously improve the user experience.
First Project with a Measurable Outcome
Thanks to Korbit being a data driven company, I was able to participate in my first project that generated a measurable outcome. My colleagues have also given me great feedback and support, which had made being the sole designer less challenging :)
What I Would've Done Differently
The KickOff meeting overran by a lot. I didn't know about the concept of First Principles, so I was presenting research findings, only to realize that I was unaware of some information that was collected prior to me joining the company. Thanks to my colleague for suggesting the use of First Principles, we were able to quickly gather everything we know that are relevant to this project. This in return, laid a foundation for our next steps.
Another issue contributing to the overrunning meeting was my failure of facilitating the HMWs brainstorming session. Instead of being lenient and letting everyone fill in their ideas (slowly), I could've set a strict timeline that forces us to come up with as much ideas as we can in e.g. 10mins (or use my beloved crazy 8s method). This will leave enough time for discussing and converging.
.png)
.png)
.svg)