See your numbers without a BI team

The problem

In most small companies, reporting is a person. Someone exports from one system, pastes into a spreadsheet, cleans it up, and produces a monthly summary that is out of date by the time it is read. When a question comes up mid-month, the answer is "I would have to pull that", which usually means the question is dropped.

Two failures follow. Decisions get made on impressions, because the number was not available at the moment it was needed. And when numbers do arrive, people argue about the figure instead of the decision, since two departments count the same thing differently and nobody wrote down which definition is right.

Website numbers have their own version of this. The default analytics tool is installed on almost every site and opened by almost nobody, because finding a simple answer in it takes longer than guessing.

What AI can actually do here

Self-service is the actual answer, and most of it is not AI.

A BI tool connects to the database where your data already lives and lets people ask their own questions, by clicking through fields and filters if they cannot write SQL, or by writing SQL if they can. The result becomes a saved question, and saved questions become a shared dashboard that is always current because it queries the source rather than a copy. That is what removes the monthly export ritual.

The AI layer helps at the edges: generating SQL from a plain-language question, suggesting a chart, summarizing what changed. Useful once the data is connected and modelled, and misleading before it, because a confident answer computed over a badly joined table looks exactly like a correct one.

For website traffic specifically, the honest fix is usually simplification rather than more capability: a dashboard with the handful of numbers that inform decisions, readable in thirty seconds.

How the tools differ

Metabase is the BI layer. It connects to more than 20 data sources, supports click-based questions and SQL side by side, and its open source edition is free with unlimited users if you can host it. Cloud Starter is $100 per month including 5 users, with Pro at $575 per month including 10.

Plausible Analytics covers website traffic, with all key stats on one page, no cookies and EU hosting. Starter is $9 per month for up to 10,000 monthly pageviews, with a 30-day trial and no free plan.

Airtable is the option when the data is not in a database yet. If the numbers live in spreadsheets, structuring them there gives you views, interfaces and reporting in one step, at $20 per user per month on Team.

When you don't need a tool

If your business has three numbers that matter, and you can get all three from tools you already pay for, write them in the same place once a week and stop. Most small companies do not have a reporting problem, they have a not-looking problem, and a BI tool does not fix that.

If your data is scattered across a CRM, an accounting tool and several spreadsheets, connecting a BI tool to that mess produces fast wrong answers. Consolidation comes first, and that is a data project rather than a purchase.

And be honest about what a dashboard changes. A number nobody acts on is a number nobody needed. Start by naming the decision you would make differently if you knew, and build only the report that informs it.