The Data Analyst Stack
Model it, ship internal tools, put dashboards in front of people, and trust the event data underneath.
You are the data function. Nobody else writes the transformations, builds the internal tool, or explains why two dashboards disagree — so the stack has to cover modelling through to delivery.
Claude
SQL & Analysis Pair
9.0/10
dbt
Transformation Layer
Metabase
BI & Self-Serve
Retool
Internal Tools
Segment
Event Pipeline
Why These 5 Tools Work Together
Fix the event layer before you buy a nicer BI tool. A beautiful dashboard on bad events is just a faster way to be wrong.
Transformations belong in version control with tests. If a metric definition lives only inside a dashboard query, it will silently fork.
Dashboards answer questions. Internal tools change things. Most teams only buy the first and then wonder why ops still runs on spreadsheets.
The combined workflow: Segment captures events into your warehouse, dbt models them into tested tables, Metabase exposes those tables for self-serve questions, Retool wraps the ones that need writes into internal tools, and Claude helps you write the SQL, review the models, and explain the results.
Who This Stack Is For
The modern analyst job is four jobs stacked in a trench coat: define the metrics, model the data, build something people can click, and answer the same question fifteen times a week. This stack maps to those four. Claude has quietly become the best pair for analytical work — it writes and explains SQL, reviews dbt models, and turns a nine-table schema into a plain-English answer for the exec who asked. dbt is the modelling layer: version-controlled transformations, tests that fail loudly, and documentation generated from the code rather than from good intentions. Metabase is the self-serve layer, because the fastest way to stop answering the same question is to let people answer it themselves — and it is genuinely open source, so the free self-hosted path is real. Retool covers the gap dashboards cannot: the internal tool where someone actually changes a record, approves a refund, or backfills an entry. Segment sits at the top, because if the events are wrong, everything downstream is confidently wrong. Budget $250-500/mo depending on seats and hosting choices, and less if you self-host Metabase.
How the Stack Fits Together
- 1. CaptureSegment
Clean, schema-checked events into the warehouse.
- 2. Modeldbt
Tested, documented, version-controlled marts.
- 3. ExploreMetabase
Self-serve dashboards on trusted tables.
- 4. ActRetool
Internal tools that write back, not just read.
- 5. ExplainClaude
SQL written, models reviewed, findings translated.
Your Stack, Tool by Tool
A breakdown of each tool, what it does in this stack, and how to get started.
Claude
SQL & Analysis Pair
Writes window functions you would have to look up, reviews dbt models for logic errors, and turns query output into an explanation a non-analyst can act on.
Key feature: Large context handles an entire schema plus the model file in one prompt.
In this stack: Paste dbt models and Metabase queries in; get back tested SQL and documentation.
dbt
Transformation Layer
SQL-based modelling with tests, lineage, and generated docs. The difference between a warehouse and a pile of tables.
Key feature: Tests that fail the build when a metric breaks, before anyone sees the dashboard.
In this stack: Reads Segment-loaded raw tables; outputs the marts Metabase and Retool query.
Metabase
BI & Self-Serve
Dashboards non-technical people will actually open, with a question builder that keeps them out of your inbox. Open source if you self-host.
Key feature: Self-hosted free tier with full query and dashboard capability.
In this stack: Points at dbt marts so every dashboard shares one metric definition.
Retool
Internal Tools
Drag-and-drop internal apps over your database and APIs. Admin panels, approval queues, and backfill tools in an afternoon.
Key feature: Write-back to production data with permissions and audit logs.
In this stack: Queries the same dbt marts; handles the actions Metabase cannot.
Segment
Event Pipeline
One tracking implementation feeding warehouse, product analytics, and marketing tools. Protocols enforce the schema before bad data lands.
Key feature: Schema violations blocked at the pipeline instead of discovered in a dashboard.
In this stack: Lands raw events in the warehouse that dbt then models.
What This Stack Costs
A clear breakdown of each tool's pricing for this persona.
| Tool | Free Plan? | Starting Price | Best Plan for Data Analysts |
|---|---|---|---|
| Claude | Yes | Free / $20/mo | Pro ($20/mo) |
| dbt | Yes | Free (Core) / $100/seat | dbt Core (free) or Cloud Team |
| Metabase | Yes | Free (self-host) / $85/mo | Self-hosted or Starter ($85/mo) |
| Retool | Yes | Free / $10/user/mo | Team ($10-50/user/mo) |
| Segment | Yes | Free / $120/mo | Team (~$120/mo) |
| Estimated total monthly cost | $140/mo (self-hosted) — $500/mo (managed) | ||
Prices as of April 2026. Most tools offer free trials.
Get Started in 30 Minutes
Sign up in this order to get the most value with the least friction.
- 1
Segment
Write the tracking plan before any code. Ten well-named events beat sixty ambiguous ones.
- 2
dbt
Start with dbt Core and three layers: staging, intermediate, marts. Add tests on every primary key from day one.
- 3
Metabase
Self-host first. Expose only mart tables — never raw — so definitions stay consistent.
- 4
Retool
Build the one internal tool that currently runs on a spreadsheet and a prayer.
- 5
Claude
Feed it your schema once, then use it for query drafting and model review rather than starting from docs.
Swap-In Alternatives
Nothing here is sacred. If one tool does not fit how you work, these are the substitutions we would make.
Metabase → Looker Studio
Free and fine for Google-native teams; weaker modelling and permissions.
Retool → Airtable Interfaces
Faster to build if the data can live in Airtable rather than a warehouse.
Segment → RudderStack
Warehouse-first and cheaper at high event volume.
Claude → ChatGPT
Better code-execution tooling; Claude reads long schemas more reliably.
Bonus Tools Worth Adding
Optional additions once the core stack is running smoothly.
n8n
FreeSelf-hostable pipeline automation for the jobs that are not transformations.
Read the n8n reviewEmail me this stack
Get The Data Analyst Stack as a plain-text checklist — tools, plans, costs, and setup order — plus a note when we update it. No gates, no spam.
Also subscribes you to The Weekly Stack. Unsubscribe anytime.
Who This Stack Isn't For
If any of these sound like you, a different Stack Pack will serve you better.
Marketers who need campaign reporting, not a modelling layer.
Check out The Marketer StackFounders pre-product-market-fit — this is premature infrastructure.
Check out The SaaS Founder Stack