A start-up client in the finance industry was struggling with user retention for their investment app. Our client wanted to serve investment projections to power-users, but without a BI system could not identify this portion of their userbase. Additionally, no system existed to actually produce and deliver these projections. User feedback showed frustration with this lack of advisor functionality, which is standard for larger providers. This meant the app’s userbase had declined by roughly 10%, threatening funding and preventing our client from growing their business. A two-stage CRM strategy was desperately needed to provide necessary user intelligence and deliver investment predictions, with a strict time limit on implementation
Our initial concern was to fully understand the data streams available to our client for power-user identification. We mapped their databases and produced a detailed data dictionary – before drafting four IDing strategies using leading cloud providers and proprietary algorithms. Our client chose to work with a fully custom solution, in order to totally own all portions of their CRM BI platform. Our data experts synthesised datapoints into a data lake on private cloud architecture. This was then fed into custom case-based reasoning AIs versioned in Python on MLflow. The next step was to design bespoke React dashboards able to visualise these new user metrics, including power-user IDs, for decision makers
The second prong of this CRM strategy necessitated an investment projection tool. Our experienced product managers consulted with contacts in the finance sector, working with leading academics in the field of AI to carefully build highly-accurate prediction models using MLflow. Hosted on private cloud, these AIs benefitted from regular training to remain relevant, handled by our clients tech team. AIs used both case-based reasoning and regression algorithms to deliver deep real-time investment insights backed by expert knowledge. Finally, our C# specialists implemented APIs connecting on-app graph visualisations to these models, delivering live insights to device-IDs associated with power-users – guiding successful investment and boosting satisfaction
