✅ Journey mapping ✅ Conversational design ✅ User testing
Challenge
Unblock the workstream, discover and optimise the tone and interaction that would resonate with British Gas/Centrica customers the most.
30k
Chat logs analysed
88
Design tickets resolved
Although the business had a NLP model in-training, I explored a tree-logic based chatbot model
Research
We conducted a mix of A/B testing, card sorting and critical use case prototypes to evaluate different design approaches.
We learned that typos, autocorrect and slang all needed to be addressed as use cases.
Repeated user testing sessions that gathered valuable feedback
Participant bias
Initially, we redacted the British Gas branding from the designs in order to get unbiased feedback from users.
Also, I innovated a data-driven approach to the UX workstreams using MS Teams Planner to track, assign and manage the workload across the team.
Initially, we redacted the British Gas branding from the designs in order to get unbiased feedback from users.
Also, I innovated a data-driven approach to the UX workstreams using MS Teams Planner to track, assign and manage the workload across the team.
MS Team Planner app, pivotal for data-driven design projects
Planner gave us team performance analytics, revealed blockage causes and automated reporting to upstream stakeholders. This way we made more progress in three weeks than in the preceding three months before I'd joined the project.
The initial flow diagram was frozen in a PDF, so I rebuilt it as a Figma file that we could control
Mental models
Codifying singular terms of reference internally for each part of the experience was key to gaining alignment and closing gaps in our squad comms.
Codifying singular terms of reference internally for each part of the experience was key to gaining alignment and closing gaps in our squad comms.
Working directly with the SAP solutions analyst to look at the data flows along with using real-time assets was instrumental in getting to a feasible set of MVP user journeys.
Nobody likes feeling in-limbo, so keep the user informed and engaged
Designs for user testing across multiple chatbot flows and use cases
Outcomes
💠 SMT approved pivoted strategy I initiated, in response to COVID-19
💠 Future-stage MVP was unblocked, tested and scoped successfully
💠 SMT approved pivoted strategy I initiated, in response to COVID-19
💠 Future-stage MVP was unblocked, tested and scoped successfully
A complex data journey with IDV issues to factor-in
Using an audio notification to mitigate latency during bot-to-agent handover
