Replacing Power BI with a local BI system: what carries over and what gets rebuilt

Short answer: Power BI reports do not move to any local BI system at the press of a button. Visualisations are rebuilt, DAX measures are rewritten, and Power Query transformations usually move into a separate data warehouse. That makes it a convenient moment to separate calculation from display: the next change of tool then costs far less. Reports are converted one at a time, starting with the ones management reads, and for a while the numbers are reconciled across both systems.

Why this comes up now

Microsoft stopped selling in Russia in 2022. Power BI Desktop still downloads and works on an individual machine, but publishing reports to the cloud service, sharing them and refreshing them on a schedule all run into licensing and payment. Companies live on workarounds for years, and any of those workarounds can stop working.

What to move to

There are several Russian BI systems, Visiology, Luxms BI, Yandex DataLens and Modus BI among them. DataLens has an open edition you can deploy on your own server. Of the open-source tools, Apache Superset is the common choice.

Choose on two questions: where the data will live, and who will build the reports. If analysts who are used to DAX build them, how close the data model and formula language are matters. If hundreds of people read them, access rights and load matter. Compare the candidates on two or three of your own real reports; on feature lists they are nearly indistinguishable.

What carries over and what does not

Data sources. Connections to the accounting system, databases and files are set up again, and that part is quick.

Power Query transformations. Cleaning, joining tables, deriving reference data. These do not move to another system, and they usually go into a data warehouse where they run on a schedule.

DAX measures. Measures and calculated columns are rewritten in the new system’s language, or moved into the warehouse as ready-made metrics. Complex measures that depend on filter context take the longest.

Visualisations. Rebuilt from scratch. A good moment to throw out the report pages nobody opens.

Why a warehouse

When the calculations live inside Power BI files, every report has its own version of revenue and margin. The move exposes that immediately: two measures with the same name produce different numbers.

Move the calculations into a warehouse, ClickHouse or PostgreSQL for instance, and the BI system only displays ready metrics. The numbers agree across reports, and the next change of BI tool touches only the visualisations. How such a warehouse is built for marketplace data is covered in what margin analytics costs.

The order of the move

List the reports and check which of them were opened in the last three months. Part of the list drops out immediately.

Start with management’s reports: there are few of them, and differences in the numbers show up there. For a few weeks both systems run side by side and the figures are reconciled. Discrepancies turn up almost every time, and more often than not the error was in the old report.

Then convert the rest department by department, switching Power BI off for the people who have already moved.

Common questions

How long does the move take? It depends on the number of reports and the complexity of the formulas. A dozen reports on clean sources convert in a few weeks. Hundreds of reports with the calculations inside the files is a project of several months, and most of the time goes on the warehouse.

Can we keep using Power BI Desktop? For personal analysis on your own machine, yes. For company-wide reporting without the cloud service and licences it is awkward and risky.

Do we need a warehouse if we only have a few reports? With a handful of reports on a single source you can do without one. Once there are several sources and metrics are calculated differently in each report, a warehouse pays for itself at the first reconciliation.

The rule

Do not port reports one for one. Move the calculations into a warehouse, convert only the reports people actually open, and reconcile the numbers before Power BI goes off.