Building a Data Stack: Series A

3 min read
Thumbnail for Blog Post - Building a Data Stack: Series A || blog/data-stack-series-a/datastack-series-a.jpeg

We talked about tips & tricks for when you raised seed — this week it’s what we think you should consider doing just after raising an A round! Lots of work to do, and finally you have budget to build out an analytics function. What should you prioritise?

  • DO ✅: MOST IMPORTANTLY: just as before, list up what business outcomes you want to improve using data. Have a clear vision of what decision are necessary to make, and how analytics can help you make them faster and better.
  • DO ✅: If you have not already, set up your data warehouse layer. We love Snowflake but BigQuery and Redshift will work equally well for your purposes (and might come with attractive credit discounts).
  • DO ✅: Populate this warehouse with raw data: use tools like Airbyte, Fivetran or Stitch to connect to third party data sources (ELT), and ensure event pipelines like Snowplow or Rudderstack to capture your users journey in your apps or websites.
  • DO ✅: This is the time to build your centralised data model. Use techniques like Domain Modelling (ask us!) or Kimball to ensure a good design from the start — this needs to be a single source of truth, and therefore you should think carefully through what you need to use it for!
  • DO ✅: Once the logical data models are built (joining different data sources together e.g. to attribute a subscription conversion to spend on a paid marketing channel), think through how you would build your presentational model to visualise this data in a clear, structured way. This is the single source of truth: all data tools consume data from this layer.
  • DO ✅: Set up the initial dashboards & reports (we typically start with Metabase and move to self-serve tools like Looker or Holistics when the data team is growing)
  • DO NOT ❌: Do not overcomplicate your data model just yet! Focus on what is truly critical to your business decision making now. You will have all the raw data to build any type of sophisticated model later on.
  • DO NOT ❌: Also don’t build reports that query raw data! Ambitions to replace them with a proper data model will almost never be realised, and the result is you end up with multiple versions of truth which will seriously jeopardise your ability to make fast decisions.
  • DO NOT ❌: Do not expect data to revolutionise your business immediately! We think it is best seen as an optimisation tool that allows you to be (much) more confident in product or marketing decisions you are making. It will NOT tell you what to do to 10x your business — that is alas still up to you.
  • DO NOT ❌: And lastly, hire carefully — do not hire a data scientist if all you need is maintenance of the above data engineering stack! Consider using fractional teams (like we offer) or hiring more analytics engineering focussed specialists.

At Tasman we have helped out dozens of growing companies in the Series A stage making the choices above — don’t hesitate to contact us for a free consultation!