Buying Analytics Platform

Buyer's guide.

Define the scope first

Product analytics (Amplitude, Mixpanel) is different from BI (Looker, Tableau, Hex) is different from data warehouse (Snowflake, BigQuery, Databricks).

Pick the layer you are buying. Buying a BI tool when you need product analytics wastes a year and a budget.

Most companies need 2 of the 3. Warehouse + product analytics, or warehouse + BI.

Warehouse selection

Snowflake, BigQuery, Databricks dominate. All three are fine for most workloads.

Pick based on: existing cloud (BigQuery on GCP, Snowflake on AWS/Azure, Databricks anywhere), team skill, data shape (Databricks for ML-heavy, Snowflake for BI-heavy).

Don't run your own ClickHouse or Postgres warehouse unless you have 100s of TB and a dedicated platform team.

BI tool selection

Looker for governed metrics layer (LookML). Tableau for self-service visual exploration. Hex for notebook-style analysis.

Modern stack: dbt models the data, Looker or a competitor exposes it. Don't skip dbt; ungoverned BI is a mess at scale.

Avoid "AI BI" tools that promise natural-language queries until they prove out at your data shape. Most still hallucinate joins.

Product analytics

Amplitude and Mixpanel are mature; PostHog is open-source and growing.

Cost scales with event volume. Plan for 5-10x event growth in year 2.

Integrate with the warehouse. Product analytics tools should backfill into the warehouse so SQL analysts can join with revenue and user data.

Buy what you'll use

Don't buy three tools when you'll use one. Each tool needs an owner, training, and integration work.

Buy 2 max in year 1. Add the third only when the gap is felt by users.

Negotiate annual deals with consumption caps. Analytics spend balloons quickly; cap it on day one.