Building a World Class Data Team
We implemented a modern data stack (Snowplow, Redshift, Looker), built the initial set of dashboards that allowed the teams to self-serve, analysed customer retention behaviour, and handed everything over to a team we helped hire.
Kaia is a companion app for people suffering from chronic back pain treatment. Patients can supplement their existing treatments with personalised exercises and get advice from professional coaches directly within the app. Their business model is subscription-based and part B2C (patients downloading and subscribing directly), part B2B2C (patients acquiring a subscription via their health insurance provider). Success is measured by the number of daily workouts patients complete, and the reduction in their self-reported pain levels. Both are particularly important for the B2B2C side of the business as the insurers want to see that Kaia can help reduce the number of visits to the therapist.
LocationMunich & New York
The team was suffering from major BI bottlenecks with outdated BI tool & silo-specific reporting (Apple App Store installs & subscriptions in one report, Google Analytics in another, Backend Report data for financial transactions, etc). We needed to find a way to solve that quickly and efficiently. There were also extra complications because of necessary compliance: Kaia is a healthcare product operating in the United States, and therefore needs to follow much tighter regulation than a 'normal' consumer app. That meant that they could not work with typical managed services (e.g. the Snowplow one), and had special requirements working with providers like Looker.
- Understand the acquisition of new customers — this is crucially important as one of the company targets was to increase the % of B2C customers significantly.
- Hire, structure, and train a team to own and build on the data platform we set up.
- The investors from the A Round (raised Q4 2018) were very keen on Kaia starting a data strategy with the end goal of informing and improving marketing spend efficacy.
Data tracking & ingestion with Snowplow
We selected Snowplow Analytics open source event data pipeline, with a support contract from Snowplow at cost as the Managed Service but with the ownership of the data and infrastructure in Kaia's hands, making it HIPAA compliant. We set up data collection in the apps, website, and backend—all running via Snowplow into Redshift. Data modelling was done with an Amazon EC2 server orchestrator, facilitating maintenance and transparency for the Kaia analytics engineers (who took over the data model after our engagement ended).
Insight delivery was always part of the initial spec, but the infrastructure needed to provide clean first before insight could be produced at scale. We set up Looker for visualisation, offering explores per data silo (Marketing, Product, Finance) and linking the silo's together for explorative analysis (for example, the impact of marketing decisions on product usage and finance). We built around 15 dashboards to serve the needs of the different teams.
We delivered a set of iPython Notebooks with two goals:
- Simple insight delivered straight away (new conversion drivers were identified that the product team were not aware of, leading them to change their onboarding flow accordingly; user segments were defined that allowed product to better understand why some groups of customers behave differently)
- But also to be used as templates for future internal Kaia data science work (as we owned the structure and modelling of all the data, this was a crucial part of the deliverables).
Building the team
No data team was present when we started the project, and when we finalised the engagement in July 2019 there were two full time data team members we handed over to. We designed and delivered a profile for Head of Data in a challenging recruitment environment (Munich). We assisted in the recruitment process in detail: not just the job spec but also the build of interview tasks, technical fit, etc. Delivery of job specs for the other team members was done as well:
- Analytics Engineering (responsible for collecting, modelling and delivering clean & structured data)
- Data Science (consuming the clean & structured data and delivering insight in the form of Looker dashboards and ad-hoc deep analysis)
We then set up a team spec and established goals for 2019—setting up the team to work off of a roadmap built at start of the project based on stakeholder interviews.
- We built the infrastructure to collect, clean, transform and model client interactions in the mobile app and on the website (Single Source of Truth).
- We built a reporting suite in Looker covering all major business metrics (giving parity to legacy reporting systems first, and then extending from there) and offered curated self-service to product managers.
- We worked with product to build a data strategy where data informs product design; and product is built with data collecting in mind.
- We helped Kaia hire an internal Head of Data, who can take over the strategy side of the project.
- We helped Kaia understand additional data team profiles, following industry best practice.
- We delivered templates for deep insight work: conversion drivers, churn drivers, customer segmentation and applied them to the available data for immediate product direction.