Tandem Bank is a UK challenger bank which offers credit cards, fixed-savings products and current accounts, all managed by its mobile app, which gives users various additional features, such as auto-saving, and aggregating non-Tandem bank account data from other accounts or credit cards.
Many of these new features have been enabled by a change in the company’s approach to its infrastructure. The company had a managed data centre offered through IBM reseller Blue Chip. It consisted of a combination of Microsoft products and SQL servers.
When Paul Clark was made CTO back in April 2017, he made the decision with senior management that on-premise systems were not exactly what was needed in the business.
“What we had wasn’t scalable, we couldn’t change it very quickly, and we couldn’t adapt to changing marketplaces or the changing needs of customers,” says Difa Niculescu, IT director at Tandem Bank.
The bank’s IT team, including Niculescu and Clark, began speaking to Amazon Web Services (AWS) around working on pilots on cloud usage, mainly around machine learning and data warehouses. The company wanted to be able to use machine learning to retrieve customer insights and personalise the way it interacted with them through its mobile app.
“AWS was selected because it was the most successful cloud vendor, which was having good traction with onboarding financial services clients. It also has the most mature platform-as-a-service (PaaS) services, which allow us to integrate quickly and adopt services in the early maturity stages of cloud,” Niculescu says.
“We started with a pilot approach where we could integrate AWS with the on-premise set-up and have some machine learning coming from our database systems, and finding the insight this way, and eventually we moved on to a conversation around moving entirely to AWS,” he adds.
After working on governance with AWS partner Nordcloud, Tandem Bank created a small core team internally, consisting of engineering staff who had been supporting the existing hybrid set up.
“They learned and trained as we went throughout the process and this was a very difficult 10 months to really rebuild the existing platform and make it cloud ready,” says Niculescu.
Working alongside Nordcloud, the ban decided to take a zero-patching approach, which meant that it had to rebuild the infrastructure as infrastructure as code (IaC).
“This makes sure that we can script and automate the build of the entire infrastructure and environment,” Niculescu says, adding that this took several months of trial and error.
Once the organisation was confident it could demonstrate that it could mimic the infrastructure and replicate it in a development environment, it worked on IaC all the way through to production.
“Zero patching is very important for us because it means we can always build the latest deployment of Windows, and the latest version of our app and our code and only build what we need which helps with speed and stability,” Niculescu states.
The build process for developers is now around 40 minutes, compared to weeks or months before taking on this approach.
“The key is the IaC piece of what we do now, as in the non-production environments, people work on them during the day and then we turn off the environments every night, saving us money. Then, they rebuild them in the morning, testing the repeatability of the infrastructure every day, and we’ve seen that we’re getting fewer bugs in the code as a result,” he says.
The project from its existing set-up to go all-in on AWS and IaC took 10 months, but the migration took one evening – starting in the afternoon and finishing in the evening.
When the company launched production, it immediately saw a 47 per cent increase in speed.
“Once we made it live, and using the app we did some testing instantly, but just using the app you could see it was much quicker,” he says, adding that consumers had noticed the changes too as ratings in the App Store have significantly increased since.
“It’s continually improving, part of that is the AWS platform and infrastructure services supporting that and part of that is good code quality and build. We now have the ability to rapidly introduce features and bring things to market, so agility and scalability has gone through the roof,” he says.
An example of this is the launch of its Autosavings feature earlier this year. It uses a number of algorithms to automatically put money aside for its customers based on their spending habits and affordability.
“This was turned around within weeks rather than months or years of development, very much on the basis of the flexibility that the developers have and the way they’re able to iterate their code and bring features to the market very quickly,” he states.