User guidance: how to provide meaningful feedback to users
How we improved the way we guide users to make them achieve a complex task
At Onfido, we’re creating an open world where identity is the key to access.
Since our goal is to prevent fraud while maintaining a great user experience for legitimate people, we launched advanced facial verification. With this new approach, we ask people to record live videos instead of static selfies. Users record themselves moving their head and talking out loud to prove they are real people.
Our clients adopted this new feature. But after a few weeks, we noticed that the videos were hard to process, making identity verification difficult. Users were not performing the right actions. They would retry multiple times and many eventually gave up. This meant that legitimate users were falling through the cracks.
We needed to find a solution, so over a Design Sprint, we talked to machine learning engineers, biometrics specialists, mobile developers, product managers and UX designers. Our goal was to radically improve our selfie video verification.
This series of blog posts will document our process to improve the selfie video feature. You can try it now on our demo app.
On part 1 of this series, we talked about adding friction to improve UX. I’d like now to highlight the importance of user feedback in the product.
A big part of the success of this new version of selfie video is the way we handle user guidance.
Recording a video of your face following actions on the screen is not easy. Many of our users have never done anything like it before. It was crucial for us to provide users with tips and feedback to guide them through the process.
The selfie video verification starts with a simple task: “position your face in the oval”. To help users to be correctly positioned, we’ve added face detection to this step. We didn’t overthink it, and the first iteration was straightforward: once a face is detected, the experience moves onto the next step.
Detecting a face is really, really fast for a machine. Fast is great! We were very happy with our face detection performances. Then, we tested it with users. Their feedback was not encouraging:
“What just happened?”
“I think there’s a glitch”
“Wow, I didn’t have time to look at that screen, what happened?”
We were trying to go too fast, users didn’t have enough time to understand what was going on. We decided to add more visual feedback to this step. We also added a small delay that gives users time to scan the screen and understand what’s happening.
When the face is detected, the next step is for users to turn the head to the side for a few seconds before looking back at the screen.
The main problem we noticed here was that users looked to the side… and stayed that way. They were peeking at the screen with the head facing the side in a very uncomfortable position. We painfully watched videos of people stuck in this position for seconds, waiting for something to happen on their phone.
We partially solved this problem with our revamped introduction screen that shows how to do the head turn correctly.
We also improved the capture experience itself to help users while they are doing the head turn.
When people turn their head, they don’t look at the screen anymore. This is a crucial moment where we can’t rely on visual feedback anymore, because they are not looking at the interface.
We added haptic feedback to the experience. In more common words: we played with phone vibrations.
We ran user testing sessions to test different vibration patterns. It was important for us to understand how people react to different types of haptic feedback.
We finally settled on a different vibration pattern for positive and negative feedback. The goal was to make users noticed something happened when they were not looking, and push them to look back at the screen where they would see a helpful message.
When users are turning their head, we’re detecting their movement to let them know how much they still need to move. This allows us to get a perfect video, ready to be analysed for fraud.
As a final step, we’ve added a playback of the video at the end of the experience.
In the previous version of this feature, users were not able to look back at the video they recorded. We made that decision because we didn’t want people to focus too much on how they look. In the end, this video won’t be public in any way. Its goal was to be a good enough selfie video that can get users verified.
We changed our minds after user research. We want to give more power to our users. They should be able decide what to submit for identity verification.
Replaying the video helped us reduce the number of failed verifications due to low-light settings, or videos where the sound was defective.
Creating a new experience users are not familiar with is a challenge. By adding friction to our identity verification experience, we took the risk to see users drop-off. By ensuring a robust process that guides them along the way, that provides feedback and control at key moments, we crafted an experience that people trust.
❤️ Thank you Rae and Sérgio for helping me making this post better.