The Cogo app helped people reduce their personal carbon footprint by analysing their spending via Open Banking. It generated a lot of data, but years of tech debt and a pivot in the app’s function had left that Insight & Impact Measurement disorganised and hard to trust, let alone analyse.
Specifying what the app should track
I started by specifying the data points the app needed to generate as users tapped and scrolled through it. The in-house developers built this tracking to the specifications. Because the app was also being used to test behaviour change theories, writing these specifications drew on my background in psychology and statistics, as well as a detailed understanding of the product itself.
Making the data usable
Once the new tracking was live, I built a suite of charts in Amplitude. The charts provided both targeted assessments, such as A/B tests comparing which features helped users reduce their footprint, and longer-term trend tracking, such as footprint reductions and app use over time. This enabled the team could prioritise what to build next with real evidence behind it.
Cleaning up the backlog
Alongside the new tracking, I took the lead on improving data quality across the board: tidying up Amplitude, where I deleted or merged thousands of old and broken events and became gatekeeper for what new data got collected; and working through how Equifax transaction data mapped to the app’s categories and its proprietary carbon model.
Every stage was documented meticulously, so for example when event names changed, the SQL queries in Metabase could be updated quickly, rather than the data science team having to rediscover what had broken when they next needed those datasets.