Fennibi Model Optimization
Model Optimization: fast Power BI reports, measured
Your report takes forever to load, and your users wait along with it. Model Optimization takes on the model behind the report, rebuilds what slows it down and measures every change. You get the changes back ready to apply, at a fixed price after an initial call.
Who it is for
- You lead a business team whose most important report takes forever to load, and your users complain.
- You built the model and want help with the rebuild instead of hunting down measure after measure yourself.
The problem
- Every page of the report takes a long time before the numbers appear. In the meeting, everyone waits together.
- Users complain, and some fall back on their own Excel extracts.
- The same code sits copied in many measures. Every change has to be repeated in many places.
- Nobody knows which measure slows things down. Rebuilding on suspicion takes time and often achieves nothing.
What you get
- A look at the data model: relationships, cardinality, calculated columns and date tables. Whatever slows things down there is listed as a recommendation in the results. Changes to the model itself can be added on request.
- Rebuilt measures: the measures with the greatest potential are rewritten, and copied code is merged in one place as user-defined functions (UDFs).
- Before/after measurement per measure: the same result, measured run time, in black and white.
- Ready to apply: a TMDL script with the rebuilt measures and UDFs, together with descriptions of the model. On the PBIX route you also get the optimized PBIX file back.
- Handover: we go through the results with you and the model developer and support you with applying them.
The scope follows your report. First come the measures that run on the slow pages, together with the measures they call. A complete refactoring of all measures in the model is available as an option in the quote. We use an AI for the rebuild. Approving it is part of the engagement.
How it works
- 1In the initial call you show us which report and which pages are slow and how they are usually queried: grouped by what, with which filters. The call is free. Without an initial call there is no quote.
- 2You provide the model by one of the three routes, most simply as a PBIX file.
- 3Fennibi prepares a potential view, at no cost to you: how many measures have potential, how much copied code they contain and what stands out in the data model.
- 4You then receive a fixed-price quote that sets both the price and the scope.
- 5Fennibi rebuilds and measures every change before and after.
- 6In the handover we go through the results, and you get the changes ready to apply.
Three routes to the model
- PBIX file under a non-disclosure agreement: the simplest route. You provide the file and get the optimizations back directly. Your reports show how the measures are queried, and measurement runs on your data.
- Export without data: you run a notebook in your own tenant that sends only the structure and statistics of the model, no row values. Measurement runs in a test environment with test data the size of your model.
- Live: you grant read access to the model, and measurement runs on the real model.
Every result states what it was measured on. Without the reports, the measurement chooses the queries itself, and the result says so.
What is not included
Model Optimization works on the model, not on the design of the report. Redesigning the report is not included. If it is not the model that slows things down but the page itself, because it carries too many visuals and generates too many queries, we tell you: it is listed as a note in the results.
Why it pays off
- Measured, not guessed: every change is measured before and after, with the same result.
- Easier maintenance: copied code then sits in one place instead of many.
- A fixed price from real numbers: the potential view shows before the quote whether the rebuild pays off.
- No subscription: one-off, at the fixed price from the quote.