We built a full data platform, deployed insight (dashboards and deep analyses) and enabled the On Deck team to manage all of the data infrastructure internally.
On Deck had scaled rapidly using an intricate collection of no- and low-code tools that individual functions implemented independently of each other. While this empowered the team to grow operationally, it resulted in messy data generated from over 400 SaaS tools that often produced conflicting metrics. There was no single view of a customer, which hampered their ability to generate business insights.
- Create a data strategy driven by business value
- Connect the team’s high-priority data systems into a reliable, centralised architecture
- Implement scalable processes that allowed On Deck to make modular technical decisions as the tech stack evolved
We worked with many different stakeholders at On Deck, and helped to build out a large data team from scratch.
Prioritisation with Story Analysis
To identify the core workstreams based on business needs and impact, we conducted interviews with stakeholders to identity the problems and requirements. Each story then went through a process of:
- Full Story exploration, developing a product brief for each potential solution to their problems
- A technical review of the Story upfront to create a costed delivery estimate which feeds through into planning
- Building a backlog of agreed Stories and Epics, with a flexible and routine cadence to reprioritised to reflect changing business needs.
Domain modelling the data
Quality control with pipeline testing
Strengthening the stack through robust quality control made sure that the new architecture could adapt and scale with changes and new needs. We took a 3-pronged approach to quality control through tools, processes, and frameworks to support the team:
Tools: We implemented a tailored data import platform, with bespoke pipelines that ingested data from no-code tools in a robust way, that sit along side SaaS approaches that are quick, cheap, and reliable for the team to manage.
Processes: We trained the team in a development approach using git, rigorous code reviews, and unit testing that caught errors before deploying them to production.
Frameworks: We implemented a dbt test framework to flag errors in the data and changes in structure before they hit production tables.
- Created a centralised data strategy that reduced reliance on external solutions and allowed for best-of-breed tech tool selection
- Implemented a reliable data pipeline that proactively flags inconsistencies and breakages
- Enabled new forms of product innovation, like machine-learning matchmaking services, with clearer decisions based on measurable customer impact
- Generated ongoing cost savings by reducing spend and reliance on other SaaS tools
- Changed the team’s culture to focus on the business value of data work.