Approach:

Uncovering user needs and business requirements

During initial research, our objectives were to uncover requirements, dependencies, and opportunities for improved data quality and efficiency. We conducted focus groups along with individual user and stakeholder interviews to understand the processes, workflow, user needs and business requirements. We created a service map of the existing process, recording the different users (whether Commercial, Assurance or Finance), the tasks they’d need to perform, and the systems that would need to work together to make those tasks happen. As part of this service map, we logged all the pain points and issues users were having, which became our basis for the MVP and was used to inform what our front-end service should look like.

Conducting research with real users in context

Ensuring an easy-to-use service that is consistent

To provide a consistent look and feel for users, we used the same design patterns as the ones we’d created for the ‘Book a Secure Move’ service and conducted usability testing with real users in order to refine our solutions over multiple iterations. During the coronavirus pandemic, we seamlessly switched to remote testing, minimising any disruption to progress.

Making data consistent

Aligning the technology stack

To align with HMPPS’s technology stack, we developed the JPC using Spring Boot and Kotlin (a JVM based language). We used Gradle to define our build file and library dependencies, Spring Security for integration with the single sign-on Auth Service (this handled the OAuth2 dance, parsed JWTs and ensure users had authorisation to use the service). We used Apache POI for parsing and creating Excel spreadsheets and provided RESTful web service integration using Spring MVC. We used Spring Data and JPA to store and retrieve data from a Postgres database and used the AWS Java SDK to retrieve reporting data from blob storage (S3). Unit tests (Junit, assert4j, Mockito) ran in the cloud platform’s Circle CI instance (alerting any issues to Slack), and Docker images were built upon success and deployed to the Kubernetes cluster (EKS).

We configured the infrastructure via code (Terraform, Helm Charts, Kubernetes Secretes, gitcrypt). Logging was via Application Insights and we enhanced this with Elastic APM for further monitoring. We used Sentry for alerting and held daily meetings to discuss logs and alerts and prioritise support issues.

From the start we ensured all code was clean and well tested. We made good use of Kotlin’s features including favouring concise, functional programming, preferring immutable objects, preventing reassignment of variables and preventing unexpected null access. This made for a codebase that was easy to reason with and less likely to contain bugs.

We worked closely with developers from HMPPS to ensure all frameworks and libraries adhered to their technology stack and cloud hosting platform, and successfully handed the service over to their developers after the delivery phase.