Designing a customer intelligence platform
Stratum predicts customers future buying behaviour to help sales teams make smarter decisions about their business.
Designing a customer intelligence platform
Sales teams at manufacturing, reseller and e-commerce businesses often struggle to make sense of a cacophony of competing data and information about their customer base. A combination of expensive kitchen-sink analytical systems and historic-facing insights make prioritising customer engagement activities reactive, slow and often misplaced.
Stratum’s customer intelligence software aims to cut to the chase with predictive mission-critical metrics that allow businesses to better invest resources and time, with their most valuable and promising customers. Heading up design at Frontier Labs, I collaborated with prospective customers and Frontier’s data science and engineering team to research, define and create the Stratum customer intelligence prototype.
Helping sales experts to make smarter decisions
Stratum was the first AI-powered software experiment from Frontier Labs and was inspired by numerous consultancy projects with B2B e-commerce companies. With common strategic and practical problems across these projects and with a proven methodology, research and development efforts focused on how we might help B2B sales managers to identify their most valuable customers and those most at risk of churn.
The challenge from a design perspective started with prioritising team focus for the prototype, as there was a myriad of opportunities to tackle. Interviewing pilot customers and reflecting on our previous market and user research, we defined user personas and pain points before re-framing the problem. Mapping the business challenges and user workflows, we created user story maps to help explore opportunities that would enable Stratum to more effectively augment decision making with data-driven insights. A simple effort/impact grid helped reduce the scope and define the design specification.
With a brief and hypothesis defined, we explored what actionable information and functionality would be required to help users when predicting their customer’s future buying behaviour. Once the information architecture was defined, we created a basic sitemap, wireframes and a prototype for testing.
We found a lot of our testing and iteration revolved around the language used within the application, partly to reduce cognitive load but often because we had incorrectly assumed some industry terms were universally known. Such as changing the name for the ‘churn’ customer metric (100%=dead, 0%=alive) to ‘Customer loyalty’ and then reversing the number logic to make it more intuitive (100%=alive, 0%=dead). I’ve found this to be a common challenge when working with AI/machine learning/data, where there is a solid scientific method that needs some level of ‘translation’ for users regarding cognitive load and/or usability. More importantly, these challenges tend to afford real opportunities in ‘augmenting’ experts with machine-driven insights.
Before moving onto the user interface design, an ‘MVP brand’ was created to establish a visual language through colour, typography and iconography. It’s important not to skip this step, as research and experimentation for the product brand identity spearheads the UI and any marketing materials (social, adverts, print collateral etc) that may follow. Importantly, it makes design decisions so much easier (and quicker to make) whilst coordinating design efforts into a more cohesive product experiment.
Market validation and piloting of Stratum is ongoing.