Metro Bank, Digital KYC onboarding

Clicks up, fraud down. Metro’s KYC magic unveiled.

An open brief.

The brief came from the Product Owner and Head of Digital banking. “Open more accounts, and make sure fraud doesn’t increase.”

With a brief like this I needed to understand the current state of the account opening journey. As well as understanding the current conversation and fraud metrics.

So we had the current journeys, but the only conversion rate data that the team had was an overall conversion rate. No one knew exactly where customers were dropping off and which sections of the application journey we should focus on. So began a two day journey through 30+ conversations with different members of the Bank.

Eventually after two days of “I don’t know,” and “Try speaking to this person,” I finally came across a team that was connected to Microsoft Dynamics who could help me understand the major pain points. All they needed was the URLs for each screen, something I could do.

This data helped us to see the largest pain point and told us 2 key pieces of information:

1. We were losing the most users was during the ID capture screens up to 43% drop off.
2. We were losing more users using a desktop device (43%) compared to mobile devices (17%).

A quick look at what we built

I know you want to see what we built. So, here’s a quick summary before you take a look at how we got there.

From initial questions…

Now that I had data, current state screens, and objectives. I had a little bit more clarity on the problem we want to solve, but a lot more questions…

Using these initial questions I created a Research plan, that documented all of the activities I could use to answer them.

…to a Research Plan

Competitor analysis

Time to look at the market to see how competitors are making their KYC slick and safe. There was a lot to learn, some companies were way ahead of the rest of the market and were using voice and video verification, some were way behind. I noted all of the strengths and opportunities and took stock of some ideas that Metro Bank could use to solve the conversion issues.

Talking to our ID and Selfie capture supplier

Jumio, our third party supplier for KYC ID scans and Selfie scans, handle the KYC for multiple incumbent and challenger banks. We set up a call with them to ask them what some of the best practices are across the industry for improving conversion rates through KYC and fraud. We learned two key things from them:

1. Photo Guidelines / Instructions for how to take a photo that doesnt get decline due to human error, such as photos with glare or blur have the greatest impact from their experience.
2. They are able to share a report that shows us how many applications are failing due to those kinds of human error mistakes. Perfect for seeing where we needed to spend more time on the guidelines and show effective our solutions are with before and after metrics.

Armed with this we now had our first feature nailed down, Photo Guidelines. We were doubly confident on this feature as a starting point for improvements to the KYC journey as it was something we had previously noted as a strength from Competitor analysis.

Onboarding a Data giant

Although we had some initial data from Microsoft Dynamics, the data was limited to chunks of the KYC screen. We needed to get more granular. Through conversations with the bank we discovered that Ratesetter a company we recently purchased were using Heap analytics to see exactly the kind of data we wanted. We onboarded them quickly and discovered exactly where we were seeing problems.

Speaking to the front line

Metro Bank luckily has a vast branch network, so I visited a couple times to talk to them about the pain points we had in our account opening process online, to see if they had any ideas. I also observe their account opening processes, and noticed something quite interesting... When the account opening specialist is filling in details and the customer no longer has to answer questions, they would talk about some of the USPs in the branch, like a Coin counting machine, or the 7 days a week opening schedule.

I instantly coined the tactic that keeps customers engaged ‘Did you knows’ and had an idea for how we could keep our customers engaged on our loading screens.

Idea synthesis

Using all of the research we had conducted, I started to conceptualise some ideas that we could use to solve the problems we were facing.

Impact Vs Effort

I took these ideas to the Product Owners and the Development team. I wanted to know what they thought from an impact and effort standpoint. By having all of the right people in the room we were able to move quickly through prioritisation and build a roadmap of features we wanted to implement and which ones to start on first. The ones that would bring as much customer value as soon as possible.

De-risking through Guerrilla testing and AI

To make sure that my ideas solved the problem and were highly intuitive I used a mix of Guerrilla testing and AI.

Guerrilla participants when prompted: “You can hold the page open, how would you take a photo of your ID document without covering any of the details” held their document like this. Let’s add that to the Photo Guidelines!

Using AI heat maps we were able to simulate a users attention. We wanted to help users to focus on the:

• Reason they made their error and how they can fix it
• Push them to a mobile device to increase the likelihood of success after a failed attempt

Iterations - Writing better error messages

I asked myself: “How do I write better error messages?” and before the times of Chat GPT, I asked Google. I found a great resource from Jenni Nadler that helped me to create actually help users take better photos of their ID and Selfies.

Users who are completing an application on a desktop/laptop device are 80% more likely to receive an error because of the quality and ergonomics of taking a picture using the webcam. I designed an Omni-channel QR code that directs users to their mobile device to complete their application faster and more accurately.

Iterations - Preventing errors with guidelines

One of the 10 Usability Heuristics from Nielson Norman Group:

“...it’s important to communicate errors to users gracefully, actionably, and clearly. However, it’s even better to prevent users from making errors in the first place...”

Iterations - Improving loading screens

What were the results again?

What did I learn in this project?