Power BI dashboards for small businesses

Python ETLs, executive dashboards and predictive models for decisions that actually get made.

I connect what you already issue and receive —invoices, the bank, the payment gateway and what you file each quarter— and out of it come sales, margin and projected cash. From integrating the data through to churn or pricing models. A dashboard the board actually looks at, not a report nobody opens.

Who it is for

• Businesses with data scattered across several systems
• Boards that want decisions made on data
• Operations whose key numbers go unmeasured

Why it ends up being needed

The conversation nearly always starts the same way: someone asks how much was sold last month and three different figures come back. The one from the ERP, the one from the finance spreadsheet and the one from the payment gateway. None of them is wrong; each counts something different and nobody has written down which.

The result is that decisions get made on instinct and justified afterwards with whichever number fits best. Not for lack of data — there is plenty — but because it sits in five places, gets stitched together by hand once a month, and by the time the report is ready it is no use for deciding anything.

What fixes that is not a good-looking dashboard. It is agreeing what each indicator means, exactly where it comes from and how often it refreshes. With that written down the dashboard is the easy part; without it, the dashboard is a new place to argue about the same figure.

What I deliver

• Data pipeline — ETL in Python or Airbyte, with a warehouse on BigQuery or Snowflake.
• Executive dashboards — Power BI or Tableau, with 5–8 key views.
• Predictive models — Churn, pricing or forecasting, as the case requires.
• Automatic alerts — Slack or email when a number crosses the threshold you set.
• Metric dictionary — Every indicator with its formula and its source, so nobody argues the number.
• Two-hour training — Your team learns to read and filter the dashboard without depending on me.

What is included

• Sales, margins and VAT due, straight from your own books
• Cash projected ninety days out
• Updated automatically, with no licences to pay for
• A private link you open from your phone

What gets decided before a line is written

• What gets measured, before how — A dashboard with thirty indicators does not get looked at. The first week is for cutting down to five or eight, each with its formula and its source written down, and for dropping the ones nobody will use to decide anything. The metrics dictionary that gets delivered is exactly that.
• How far back the history goes — Pulling in ten years of data multiplies the cleaning work and hardly ever changes a decision. How much history comes in is settled at the start, and it is settled by looking at which questions have to be answerable, not at how much happens to be stored.
• How often it refreshes — A dashboard refreshing every hour costs more to maintain than one refreshing overnight, and for most decisions it makes no difference at all. The frequency is chosen by the pace at which the decision is taken, not by what looks impressive in a demo.
• Dashboard or predictive model — Almost always the dashboard first. A churn or pricing model trained on dirty data, with no shared definition of each indicator, predicts the noise — and does it very confidently, which is the dangerous part.

What is not included

• Power BI or Tableau licences, which are contracted in your name.
• Fixing the data in the source system: what arrives broken is documented, correcting it is separate work.
• Keying in by hand data that only exists on paper.
• The business interpretation of what comes out, which is your decision.

What you need to have

• Read access to the systems where the data lives.
• Someone who can decide what each indicator means when there is doubt.
• Knowing who will look at the dashboard, and to decide what.
• If spreadsheets are in the loop: who maintains them and on what basis.

Frequently asked questions

Does this work if my data sits in five different places?

That is the normal case: a spreadsheet, the ERP, the payment gateway and the CRM. The first part of the job is bringing them together and automating that process.

Who maintains the dashboard afterwards?

The pipeline is versioned and documented, so your team can maintain it. If there is no technical team, maintenance is contracted separately.

Dashboard or predictive model?

Almost always the dashboard first. Without clean data and a shared definition of each indicator, a predictive model predicts the noise.

When do I see the first dashboard?

The first dashboard is ready in three weeks, after the week spent defining metrics. From there it evolves with whatever questions come up.

Do I need a data warehouse already in place?

No. If there is none, it is set up as part of the work. If you already have BigQuery, Snowflake or something equivalent, that gets used instead of duplicating it.

My data is messy. Does this still work?

That is the usual starting point. Cleaning is part of the work, and whatever arrives broken from the source is documented so it can be fixed where it is generated rather than patched every month.