Reporting and dashboards
Reporting leadership runs on, in Power BI, Tableau, Streamlit or a custom build: models that reconcile and dashboards built with the people who use them.
Most reporting fails quietly. A dashboard is built fast for one meeting, nobody owns it, two reports disagree, and the fix is a conversation rather than a test. Within a year the people who read it have stopped trusting it.
We build reporting the other way round: a semantic model with tests and documentation underneath, dashboards designed with the people who make the roster and the budget, and a reconciliation back to source so a number on a screen is a number someone can defend. More than fifty dashboards built this way ran a 3,000+ FTE operation week to week.
What you get
Delivered, not recommended
- A semantic model with tests, documentation and lineage, on dbt or the platform's native layer, feeding Power BI, Tableau, Streamlit or a custom web app.
- Dashboards built with the people who use them and reviewed with them until they are used.
- Reconciliation back to source, so every headline number can be traced and defended.
- A refresh that runs on a schedule and flags when it fails, rather than a report someone remembers to re-run.
From the work
50+
Dashboards shipped into production
Reporting infrastructure that a function relied on
New Zealand government · same engagement
An operational function needed reporting it could run on week to week, not proofs of concept that stalled after a demo.
The reporting became the infrastructure that the function relied on. Our founder scoped, built, reviewed and maintained it throughout the five years.
Read the noteQuestions
Asked before the first call
Can you rebuild reporting we already have?
Yes, and that is the most common shape of this work. We start with the reports people rely on, put a tested model underneath them, reconcile the old and new numbers, and retire the old versions once the differences are explained.
Which reporting tools do you work in?
Power BI is where most of our reporting has shipped, and we also build in Tableau, in Streamlit, and as custom web apps when a dashboard tool is the wrong shape for the job. The modelling underneath is the same whichever one sits on top, so the choice is made on what your operation already reads and what the work needs, not on what we prefer.
Will our analysts be able to change it afterwards?
That is the point of the model and the documentation. Handover includes working sessions with the people who will own the reporting, and the semantic layer is built so a new measure is an addition, not a rebuild.