Blog - The evolution of the startup data stack

[ivanagas]

Strapline

Explain the idea in a sentence or two

Detail the startup data stack evolution from MVP to Series B based on Luke’s research.

Why should we do it?

how it will it be useful or interesting for readers/viewers

Luke’s done a bunch of research on this. It is a decision startups have to make where resources don’t exist. Targets our ICP.

Headlines options

• Startup data stack evolution from MVP to Series B+
• The evolution of the startup data stack (MVP to Series B)
• Data data stack evolution 🕺

Outline

Bullet point outline of structure / questions / topics to be covered

• What is a data stack?
  • The software, technologies, products teams use to understand their product. They use the data to understand how they are doing and decide what they should do.
• MVP
  • Get something up quick, basic, fast, proven.
  • Gives the basic knowledge they want: are people visiting, using the product, paying. Report on KPIs.
  • Do do this, they need:
    • Basic website analytics
    • Basic product analytics
    • Revenue
    • CRM
  • Just follows what others are doing. Use SaaS, cloud.
• Seed
  • More advanced tracking, likely connected to ads and multiple sources like advanced CRM, Zendesk, Facebook, Instagram, Google Ads.
  • Further analysis into product usage with session recordings, specialized data, experimentation, ROI on ads.
  • Potentially use CDP to manage all this data.
  • Decisions about how to scale are made.
• Series A
  • People specifically responsible for data are brought in. Implements the “modern data stack”
  • Data amounts start getting large, need data warehouse to handle it. Use Snowflake, BigQuery, or RedShift.
  • More data, more sources. Customer engagement platforms, more complex data.
  • Revenue and pipeline is brought into warehouse using ETL like fivetran, stych, airbyte to connect sources, like revenue, marketing, and pipeline. Clean and make useable in further processing.
  • DBT, mode to process and transform data in your warehouse.
  • Metabase to query and have nicer visualizations. BI tools like Hex.
• Series B and Beyond
  • Data amounts start getting large, need data warehouse to handle it. Snowflake becomes source of truth. Likely need to use SQL to handle it.
  • Becomes more siloed because of need to use specialized tools and workflows to access and process the data into something useable. Need to write SQL to get at data.
  • Non-data team is fully reliant on either a visualization tool or asking data team for insights. Less self-serve for engineers.
  • Product analytics becomes less of a focus. Data continues to stream in, removing or changing it is difficult.
  • Bigger BI tools, Looker. Larger data and analysis team. Seen as needing to empower engineering team with data.
  • Machine learning becomes more important.

[andyvan-ph]

I like this idea and I think it we should go pretty deep here. Would make a great lead for a future newsletter.

Things we'll need:

• Some nice diagrams from Lottie

• I think we should have a section explicitly about "our vision" for the future of the data stack – i.e. this is where we are now, and here's where we're going and how PostHog fits into that. High-level is fine, but would make a nice conclusion.

[lharries]

Great to see.

I wonder if "Evolution" is the right name for the title? I'd expect an article about data stack evolution to show what a series b sized company's data stack looked like 10 years ago vs now. However, instead we are talking about how a startup's data stack grows.

Some other ideas:

• "The modern data stack from seed to Series B"
• "The startup data stack from seed to Series B"
• "The ultimate guide to the startup data stack"

I think we should have a section explicitly about "our vision" for the future of the data stack – i.e. this is where we are now, and here's where we're going and how PostHog fits into that. High-level is fine, but would make a nice conclusion.

+1 on this. Or in each section show an alternative using PostHog to bundle the different tools?

[ivanagas]

"How the startup data stack evolves, what we’ve found" might be a good title