Which data platform?

Warehouses, lakes and lakehouses overlap more every year, and data mesh is a way of organizing people rather than a technology. Answer seven questions about your data, your users and your organization to see where to start.

1 of 7What kinds of data do you need to analyze?

Mostly structured business data from applicationsA mix of structured, semi-structured (JSON, events) and filesLargely unstructured: documents, images, logs, audio

2 of 7Who will use the data?

Analysts and dashboards (SQL and BI)Data scientists and ML engineersBoth, from the same data

3 of 7Does your data change after it arrives (updates, deletes, corrections)?

Rarely; it is mostly append-onlyOften, for example orders and accounts that change statusWe must also delete personal data on request

4 of 7How fresh must the data be?

Daily is fineWithin an hourWithin minutes, streamed from source systems

5 of 7How much data do you expect within three years?

Up to a few terabytesTens to hundreds of terabytesPetabytes

6 of 7How is data work organized?

One central data teamA central team, but it is becoming a bottleneckMany domain teams already own and publish their data

7 of 7How important is avoiding lock-in to one vendor or engine?

Not important; we want one managed serviceSomewhat importantVery important; we want open formats

Show my recommendation

---

Compare the options in detail in data lake vs lakehouse, or see how Uber moved from a day-old data lake to minutes.