Business intelligence has a pricing problem for Indonesian SMEs. The polished commercial stacks assume a budget — and a data-team headcount — that a growing local company simply doesn't have yet. But the need is real: leadership wants to see the numbers, not wait a week for a spreadsheet. Here's a BI stack that runs on one modest VPS and a week of setup, entirely on open source.

The four pieces

  • PostgreSQL as the warehouse. For SME data volumes, a well-indexed Postgres is not a compromise — it's plenty, and everyone can hire for it.
  • An ETL step to pull from the operational systems (ERP, POS, spreadsheets) on a schedule. This can be as humble as scheduled SQL and Python; it does not need to be a platform.
  • Metabase as the dashboard layer. Non-technical staff can build their own charts with the question builder, which is the feature that actually gets BI adopted.
  • A reverse proxy + tunnel so the dashboards are reachable and authenticated without exposing the database.

The design that keeps it cheap

The trick is to separate the operational database from the analytical one. You never point dashboards at the live ERP — one heavy query at month-end could stall the business. Instead, ETL lands a clean copy in the reporting Postgres overnight, and every dashboard reads from there. Slow queries hurt no one who's trying to make a sale.

Separate the database people sell from the database people report on. It's the single decision that keeps a one-server BI stack both cheap and safe.

Model a little before you visualize

Resist the urge to chart raw tables. A handful of well-named views — daily sales, inventory by location, receivables aging — turn confusing joins into obvious building blocks, so a manager builds the report they want without a ticket to IT. The modeling layer is where BI projects quietly succeed or fail.

What it costs

A small server, a domain, and roughly a week to wire it together and load the first datasets. After that, the marginal cost of a new dashboard is minutes, not a license. It won't out-feature Tableau or Snowflake — but for a company that just needs to see clearly, it's the difference between having BI and dreaming about it.