Improving Onboarding

Conversion up 29% on iOS

👨‍🎨 Product Desinger

⏳ March 2018 - May 2018

Frustrated participant


~30% of users were dropping off directly after signing up, let alone experienced the product in any way.

Before I joined Lingvist, I was eager to understand the product, even if understanding a foreign language really isn't one of my greater strengths. Quite quickly, I got a sense of the pain points. I also observed friends and colleagues trying it out, too. I noticed immediately people struggling to understand how to interact with the first card and questioned how they would learn. This wasn’t a great sign, so I went about fixing it.


Onboarding wasn’t exactly the top priority when I joined but after some gentle persuasion, I was given a budget to conduct some usability testing on new users. I invited colleagues to participate by taking notes during the sessions and recorded everything using Lookback. This way I could create a highlights reel to show the Exec team, and during our All Hands.

A lot of usability issues were discovered during the sessions, especially after sign up - 5 out of 5 candidates all struggled to complete the first card and in one instance became visibly frustrated and very angry, before I rectified the situation. It was tough to see people struggling with our product but it was a real awakening for the company.

User interviews

I also started to interview some of our existing users to get a better understanding of who they are, how they learn using Lingvist, pain points and when they reached the elusive "Aha!" moment.

Equipped with enough data - qualitative and quantitative, I created a “Learner Journey Map”. The map would look at the current onboarding flow, goals we could measure against, look at messages outside of the product and pain points.

User journey map and user persona

First problem to solve

We started at the obvious place - drop off after signing up. It’s better to build something around users’ expectations and mental models rather than forcing them to learn the system you’ve built. This was a tricky hurdle for us internally, as we want to make human learning faster. To achieve this the algorithm is designed to learn and adapt to your learning so users are learning and being tested at the same time, essentially.

This was difficult to communicate to our users and the initial onboarding process at that time was overlooked by our users.

After looking at other products and experimenting with a few ideas that didn’t work out, I went back to the drawing board. I thought about the different user scenarios of entering our product for the first time - same situation, different user personas:

  1. I know it
  2. I think I know it
  3. I don’t know it
Sketches and use cases spreadsheet
Rough sketches that didn't work out. Excel to the rescue.


This opened up my mind to start by asking the user if they know the word. If they say they know it, we show them where to type it. If they don’t know it, we can show them how to reveal the answer first and then type it.

Design flow
I know it flow.
Design flow
I don't know it.


We built the new onboarding flow on our Web app and conducted another usability test. This time 5 out of 5 participants completed the first card without any issues.

Then we released the new design as an AB test and overall the results were positive, especially for iOS.

For iOS, we saw a small 5% increase in Day 2 retention and a big positive increase in payment metrics (conversion up 29%, ARPU up 55%, ARPPU up 22%).

For Android, we saw a good 10% increase in Day 2 retention and a 10% increase in trial activations. However this increase in trial activations failed to funnel through to payment metrics - no changes there.

Find out about my biggest achievement to date