Dashboard or monthly report: the question before the tool question

It is the seventh working day of the month. The close is done, the figures are checked, and the reporting is still to come: the same twelve pages as last month, to the same nine people — management, the board, the bank, the fiduciary. At some point somebody says it should “finally be done in Power BI”. The suggestion is not wrong. It just answers a different question from the one on the table.

A report is a document, a model is a platform

The question is not which tool is better, but what the report is there for. Something the same eight to twelve people read every month, that always looks the same and has to stay findable for ten years, is a document. A data model that a dozen people filter and adjust for themselves is a platform. Microsoft describes the purpose of its own service in exactly those terms: the content is not static, you take it apart and ask questions in your own words (Microsoft 2025a). In many finance departments the job is a different one: nine people are meant to read the same page, and nobody is meant to change anything.

And the comparison is rarely with a blank sheet anyway, but with the workbook that already exists. Research disagrees about how error-prone that is: Panko found errors in 94 per cent of 85 audits of spreadsheets in real use (Panko 2015); a team led by Powell considered that figure unreliable, measured 0.8 to 1.8 per cent of formula cells in error themselves, and still found confirmed errors in 16 of 25 spreadsheets (Powell, Baker and Lawson 2009). Both figures point the same way: manual work in spreadsheets produces errors, often enough that moving away from the manual work usually pays for itself. Neither of them says what to move it to.

«The question is not which tool is better, but whether what you end up with should be a document or a platform.»

Reading is what needs a licence

In proposals this point tends to sit well towards the back: it concerns the reading, not the building. A free licence lets a person create reports only for themselves. And anyone who wants to open a report somebody else has published needs a Pro licence of their own where there is no Premium capacity — the person reading, that is, not just the person building (Microsoft 2026a). So there are two kinds of cost: licences per person, and capacity per environment.

The second path does not settle that by itself. With Premium Per User you share only with people who also have PPU, unless the content sits in a workspace on Premium capacity — and a PPU workspace is expressly not the same thing as a workspace on Premium capacity (Microsoft 2026a). The documentation puts the threshold precisely: only from capacity size F64 upwards can people with a free licence and a viewer role look at content; below that, everyone who reads needs Pro, PPU or a trial licence (Microsoft 2026b).

Then there are the people who do not sit in the house. External readers get access through an invitation as a guest account in your directory and have to sign in; private addresses then count as a social identity and cannot use the mobile apps, and only someone who holds Pro or PPU themselves may issue the invitation (Microsoft 2026c). There is one route that needs no user licence for reading: on Power BI Report Server only publishing needs a Pro licence — the server itself is licensed by cores, through a reserved Fabric instance from F64 upwards, or through SQL Server (Microsoft 2025b). So do not cost out the building; cost out the reader list: how many names are on it, and how many of them only read?

A file that stays a file

The service will give you a document too: reports can be exported as PDF, one PDF page per report page, within documented limits — no reports over 50 pages, none over 250 MB, and anyone who receives a shared dashboard from outside cannot export the reports behind it as PDF (Microsoft 2026d). An ordinary report also outputs only the rows on screen; a paginated report puts every row of a long table on paper (Microsoft 2026e).

Why that counts in accounting is written into the law. Books of account and accounting vouchers have to be kept for ten years, counted from the end of the financial year, and may be held electronically as long as they can be made legible again at any time (OR 1911, Art. 958f). The ordinance on the keeping of business records asks for no particular software, but for integrity, availability and legibility within a reasonable period; migration is allowed, but has to be logged (GeBüV 2002, Art. 3, 6, 9 and 10).

A single file is a plain way to meet exactly those requirements. How an HTML file is read is laid down in an open standard — including for documents that are not syntactically correct (WHATWG 2026). And anyone is free to implement that standard: the W3C runs its recommendations under a patent policy that requires royalty-free commitments for implementation (W3C 2025). A file like that opens in any browser, can be mailed and printed, and needs no account to read.

What the upkeep costs when the chart of accounts changes

Both paths deserve honesty here: neither of them is free. A new account, a new cost centre, a changed structure — that hits the data model just as it hits the report you had built, and in both cases somebody has to get their hands in. So when you compare a proposal, ask not only what building costs but what changing costs.

On the platform side there is the running of it. Eight scheduled refreshes a day on shared capacity and 48 on Premium, PPU or Fabric are ample for a monthly report. What matters more is the route to the data: if the source sits in the house, every scheduled refresh needs a gateway, and a model can use only one gateway connection (Microsoft 2026f).

The document side costs something as well: anyone who wants to change a file needs somebody who can do it. If that somebody is only a supplier, and nothing was handed over, you do not have a file — you have a dependency.

When Power BI is the right answer

There are situations in which the platform is the better buy, and they are more common than this piece has made them sound so far. Power BI is a mature product: a large body of users, training you can buy in, and documentation that answers questions instead of leaving them open — every limit named above can be looked up there. Anyone who recognises their own situation in the points below need look no further.

• Your readers want to filter and follow up themselves, not just read. A fixed document is the wrong form for that — and Microsoft describes the service expressly for this case (Microsoft 2025a).
• Different parts of the business are meant to see different cuts of the same data. Row-level security is the clean solution for that: one model, roles with filter rules, assignment in the service (Microsoft 2025c).
• The volume of data is large and the figures have to stay current as you go. Incremental refresh lets models grow to potentially billions of rows without everything being loaded again each time (Microsoft 2025d).
• There is already a data warehouse or a lakehouse, and a team that leads the data architecture. Then the existing landscape decides in any case, not the monthly report.
• Microsoft 365 is already running in the house, and user administration, auditing and access are meant to stay where they are already settled.
• There is somebody in the house who enjoys building models and still wants to look after them in two years' time. In our experience that is the point that decides it in the end.

Server location rarely carries the decision. Microsoft lists Switzerland North and Switzerland West as regions where Power BI and the Fabric workloads are available (Microsoft 2026g). And in law Switzerland does not ask for the postcode: disclosure abroad is permissible where the Federal Council has established adequate protection, or where suitable safeguards are in place (DSG 2020, Art. 16); the federal data protection commissioner leaves the responsibility with the company itself in any case (EDÖB 2025).

Six questions that carry the decision

If the tool question is the wrong first question, this list helps — and it can be worked through in a single meeting:

• Who reads the report, and how many of them only read? The licence question hangs on the number of readers.
• Does the report leave the house? For a bank or an auditor, a file is a different matter from access with a sign-in.
• Does the report look the same every month, or does each reader ask their own questions? Always the same means a document; own questions mean a model.
• Where is the report in year ten? If the answer is “in the year-end close file”, you need something that can be filed.
• What happens in the month your chart of accounts changes? Who makes the adjustment, how long does it take, and is it in writing?
• Who can change the report if your supplier is no longer there? An answer containing only one name is not an answer.

Where Skopa stands

Skopa does not replace the ERP and does not want to. Your system of record stays where it is. The chain starts with the export out of it, runs through the report and ends in the filing — that is the layer that gets automated, as a rule as a single document that can be sent, printed and filed with the year-end papers. The approval stays with a person, and how deep it goes is your decision.

It is built to be handed over: at the end, the reports, the template and the documented route belong to you. And if the review shows that your situation calls for a platform, Power BI is a good answer. It just belongs at the end of the question, not at the beginning.

Power BI and Microsoft are trademarks of the Microsoft group of companies. This article reproduces no Microsoft content and is not affiliated with Microsoft.