Buying Analytics Platform

Buyer's guide.

Define the scope first

Analytics platform buying is the discipline of separating the three layers. Product analytics, BI, and data warehouse each solve a different problem; conflating them is how teams end up with three half-rolled-out tools.

• Three distinct layers. Product analytics (Amplitude, Mixpanel), BI (Looker, Tableau, Hex), and data warehouse (Snowflake, BigQuery, Databricks) per need. Different categories with different vendors.
• Pick the layer first. Named-need scope per buyer. Buying a BI tool when the team needs product analytics wastes a year and a budget; clarify the actual question first.
• Most companies need two. Warehouse-plus-product or warehouse-plus-BI mix per org. Three is rare; usually a sign of an unclear question rather than a multi-layer answer.
• Documented driver. Named use case per decision. Catches solution-shopping when the vendor pitch is more compelling than the actual gap.

Warehouse selection

Warehouse selection is the foundation. Snowflake, BigQuery, and Databricks each fit a different shape; the cloud, the existing skill set, and the workload mix together point to the answer.

• Three dominant options. Snowflake, BigQuery, Databricks per vendor. All three are fine for most workloads; the differentiator is fit, not raw capability.
• Pick by cloud, skill, data shape. BigQuery on GCP, Snowflake on AWS or Azure, Databricks anywhere. Databricks for ML-heavy workloads, Snowflake for BI-heavy.
• Do not self-host under 100s of TB. No-self-host rule per org at small scale. ClickHouse or Postgres warehouse needs a dedicated platform team; the operating burden is real.
• IaC source. Terraform definition per warehouse. Supports change review and rollback when the warehouse configuration drifts from intent.

BI tool selection

BI tool selection runs by use case. Governed metrics, self-service exploration, and notebook analysis each map to a different vendor strength.

• Looker. Governed metrics layer (LookML) per org. Right when consistency across dashboards matters more than ad-hoc speed.
• Tableau. Self-service visual exploration per team. Right when ad-hoc viz dominates and the team trusts analysts to define metrics.
• Hex. Notebook-style analysis per team. Right when the output is narrative reports rather than always-on dashboards.
• Modern stack plus AI BI caveat. dbt models the data, the BI tool exposes it. Do not skip dbt; ungoverned BI is a mess at scale. Vet AI-BI natural-language-query claims with proof; most still hallucinate joins.

Product analytics

Product analytics is its own category. Amplitude, Mixpanel, and PostHog dominate; cost scales with event volume rather than seat count, which changes the negotiation.

• Mature options. Amplitude, Mixpanel per vendor. Established; broad community; predictable feature set.
• Open-source option. PostHog per org. Growing fast and self-hostable; good fit when data residency or cost prediction matters.
• Cost scales with events. Five to ten times event growth in year two per deal. Cap commitments accordingly; shipping more telemetry compounds with usage growth.
• Warehouse integration. Warehouse backfill per tool. SQL analysts join product events with revenue and user data; the integration is the lever for cross-team value.

Buy what you'll use

Apply the discipline by buying minimal. Each tool has an owner, training, and integration cost; the third tool's incremental value is usually less than its operational tax.

• Do not over-buy. No-three-when-one rule per org. Each tool needs an owner, training plan, and integration work; under-using three tools is worse than over-using one.
• Buy two max year one. Bounded purchase per year. Add the third only when the gap is felt by users rather than imagined by procurement.
• Annual deals with caps. Consumption ceiling per deal. Analytics spend balloons quickly; cap it on day one rather than asking forgiveness in the next quarterly review.
• Named owner. Responsible team per tool. Supports operational reviews and prevents the tool from drifting into orphan-status during reorgs.