Tableau to Power BI Migration: What It Costs, What Breaks, and What Never Converts

Short answer: a Tableau to Power BI migration saves very little on licences and costs real weeks of rebuild work. For 25 people, Power BI Pro at $14.00 per user per month paid yearly is $4,200 a year against Tableau Standard's published floor of $4,500, a gap of about $300. There is no file-level converter from .twbx to .pbix, because the two formats describe different things, so every dashboard worth keeping gets rebuilt by a person. Both prices were read from the vendors' own pricing pages on 31 August 2026.

That is not an argument against migrating. It is an argument against writing the business case around the seat price, which is what most of these projects do, and then discovering in month three that the money went somewhere else entirely.

Can you convert a Tableau workbook to Power BI?

Not as a file conversion, and the reason is structural rather than a gap that tooling will close next year. A .twbx is a zip archive containing a .twb workbook definition in XML plus any extracts and images that workbook needs. Inside it, every worksheet binds to whichever data source it wants, so a single Tableau workbook can sit on several unrelated connections at once. A .pbix holds exactly one semantic model and a report definition on top of it.

So a Tableau workbook spanning five unrelated connections is not one file to convert. It is five Power BI semantic models, plus a decision about how to split the report across them, and that decision is a modelling judgment about your business. No converter can make it, which is why the migration services on the market all advertise partial automation and none of them advertise a finished dashboard.

What actually breaks

Four things, in roughly the order they cause trouble.

Calculated fields. Tableau calculations evaluate in the context of a worksheet. DAX evaluates against a model with defined relationships and a filter context. Simple arithmetic ports fine. Level of detail expressions and table calculations do not, because they encode an assumption about the shape of the view they were written in, and somebody has to work out what each one was actually trying to measure before it can be written again.

Row-level security. Tableau user filters and entitlement tables become RLS roles with DAX filter expressions on the Power BI semantic model. This is the failure that surfaces latest and hurts most, because a dashboard with broken RLS still looks perfect in testing. It just shows the regional manager the whole company.

Data source coverage. Tableau ships connectors that Power BI does not, and the reverse is also true. Where a connector is missing, the fix is usually to stop connecting the BI tool directly to the source at all and land the data in your warehouse first, which is what a data integration platform is for. That is often the right architecture anyway, but it is a separate project and it belongs in the migration plan rather than in a surprise.

Extract refresh schedules. Tableau extracts and Power BI dataset refreshes have different limits, different scheduling models and different capacity implications. On Pro seats you get a fixed number of scheduled refreshes per dataset per day, which is frequently fewer than the team was used to.

What the licence swap actually saves

Here is the whole financial argument, at published list prices for the same 25-person team, verified 31 August 2026:

Two caveats on that table, both of which matter. The Tableau figure is a floor, not a quote: Tableau's pricing page lists Standard at $15, Enterprise at $35 and Tableau Next at $40 per user per month, every one labelled "starting at", and it no longer publishes Creator, Explorer or Viewer prices at all. The Creator $75 figure that appears in nearly every comparison article is reseller-reported, not something Tableau publishes. Our Tableau pricing page goes through what that means at renewal.

The second caveat is Copilot. Microsoft's documentation states that a Fabric Copilot capacity "must reside on at least an F2 or P1 SKU", and Copilot usage from Pro and Premium Per User workspaces bills to that capacity. At the $0.18 per capacity unit hour published in the Azure retail price API, that is $262.80 a month whether anyone uses it or not, which is $3,153.60 a year. If the AI features are part of why you are moving, Power BI stops being the cheap option. The full working is on our Power BI pricing page.

The costs nobody puts in the business case

The rebuild is the big one, and it is labour, so it is priced in whatever a data analyst costs you. Beyond that, three recurring items show up in every migration that goes long:

Parallel running. Both platforms stay licensed and both keep hitting your warehouse while people are being moved across. Three months of parallel running on the 25-person example above is roughly $2,175 in duplicated licences, plus double the extract and refresh compute against Snowflake or BigQuery for the whole period. On a consumption-priced warehouse that second number is often larger than the first.

Retraining. Every analyst who was fluent in Tableau calculations is a beginner in DAX for a while. This is a real and temporary drop in output that nobody schedules.

The long tail you decided to keep. Migration projects overrun because somebody promises that nothing will be lost. That promise is what turns a six-week rebuild into two quarters.

The inventory step that saves the most money

Before anything else, pull the usage statistics out of Tableau and sort every workbook by how many distinct people opened it in the last 90 days. Most estates have a long tail of workbooks that nobody has viewed in a year, built for a question somebody asked once, and rebuilding those is pure waste.

Then split the survivors into two piles. The first is the genuine recurring reports: the numbers a team looks at every Monday, which deserve a rebuilt dashboard. The second is everything that exists because someone needed a number once and the only way to get it was to ask for a dashboard. That second pile is usually larger than the first, and it does not need Power BI at all. It is ad hoc reporting, and rebuilding it as dashboards is how you end up migrating the same problem into a new tool.

Migrating to QuickSight or Looker Studio instead

Power BI is the default answer, not the only one. If your warehouse is Redshift or your stack is AWS-native, Amazon QuickSight is worth pricing: $3 per Reader per month and $24 per Author makes read-heavy access much cheaper, though a $250 per month per-account infrastructure fee triggers the moment any Pro user or Q generative feature is enabled. If your reporting is mostly on Google data, Looker Studio is free for both creators and viewers with no seat cap, and Looker Studio Pro is $9 per user per project, which multiplies by projects rather than just by people.

None of them converts a .twbx either. The rebuild cost is identical across all three, which means the destination should be chosen on where your data already lives and who you can hire, not on which vendor claims the smoothest migration. The full side-by-side is in our Tableau alternatives comparison.

One thing to check before you start: does anyone use a Mac?

Power BI Desktop is Windows only. Mac users need a virtual machine, a Windows machine, or the browser-based service, which cannot do everything Desktop can. Tableau Desktop runs natively on macOS. On a team with several Mac-based analysts this is not a footnote, it is a hardware line item and a daily friction, and it is the single most common reason a Power BI migration gets reversed.

The option that is not a migration

If the honest reason you are leaving Tableau is that the dashboards get built and then nobody opens them, moving to a different dashboard tool changes the invoice and nothing else. The workflow is the same: connect, model, build, publish, wait for someone to look.

Agentsql is the other shape. It connects read-only to the PostgreSQL, MySQL, Snowflake or BigQuery database you already run, reads the live schema so it uses your real table and column names, turns a plain-English question into SQL written for that engine's own dialect, runs it, and returns the number, a chart and the exact query it wrote. There is nothing to build before the first question and no authoring seat to buy. For the long tail of one-off questions that filled your Tableau estate with workbooks nobody reopened, that is a better fit than a rebuild. For the Monday morning numbers a whole department reports against, build the dashboard properly in whichever platform you picked. Most companies need both, and the mistake is paying dashboard prices for the half that was never a dashboard.