Engineering notes

Deep, defensible technical breakdowns on Power BI semantic model architecture, DAX evaluation mechanics, VertiPaq memory optimization, and modern developer workflows.

NOTE_01 // FABRIC & MODERNIZATION
2026-10-11·11 min read

Power BI Licensing Explained: Pro vs Premium Per User vs Fabric Capacity (2026)

What each licence costs, what it unlocks, who needs one, and the viewer count where capacity becomes cheaper than per-user licences.

›Power BI Pro costs $14 per user per month and Premium Per User (PPU) $24, both billed annually. Fabric capacity is billed per capacity, not per user.
›Pro limits a semantic model to 1 GB and 8 scheduled refreshes a day. PPU raises that to 100 GB and 48 refreshes.
NOTE_02 // FABRIC & MODERNIZATION
2026-10-11·9 min read

Microsoft Fabric Capacity Pricing: Every F SKU, Pay-as-You-Go vs Reserved

Monthly cost for every F SKU, what a reservation saves, what the free Fabric licence does and does not cover, and how to avoid paying for capacity you do not use.

›Fabric capacity is priced per capacity unit (CU): $0.18 per CU-hour pay-as-you-go, or $938 per CU per year reserved (US East).
›A one-year reservation saves about 40.5% — but only if the capacity runs most of the month.
NOTE_03 // ARCHITECTURE
2026-10-11·9 min read

Power BI Embedded Pricing: A SKUs vs Fabric F SKUs vs Per-User Licences

What it costs to show Power BI reports inside your own app, which SKU fits which embedding scenario, and when an F SKU is cheaper than an A SKU.

›There are two embedding scenarios: embed for your customers (your app owns the data) and embed for your organisation (users sign in with their own accounts). They have different licence rules.
›When embedding for customers, end users need no Power BI licence; you pay for capacity (an A SKU or an F SKU).
NOTE_04 // SEMANTIC MODELS
2026-10-11·11 min read

Row-Level Security in Power BI: Static, Dynamic (USERPRINCIPALNAME) and Manager Hierarchies

How to set up RLS that scales, the dynamic USERPRINCIPALNAME pattern, manager-sees-team hierarchies, why page-level security is not security, and how to test it properly.

›RLS only restricts users with the Viewer role. Workspace Admins, Members and Contributors see all data.
›Dynamic RLS uses one role and a security table filtered by USERPRINCIPALNAME(), instead of one role per region or department.
NOTE_05 // DAX & PERFORMANCE
2026-10-11·9 min read

How to Use Performance Analyzer in Power BI (and What “Other” Really Means)

A repeatable way to find the slow visual on a report page, read each timing correctly, and decide whether the fix belongs in DAX, the model or the report design.

›Performance Analyzer splits each visual's load time into DAX query, Direct query, Visual display, Other and (for field parameters) Evaluated parameters.
›A high DAX query time points at the measure or model. A high Visual display time points at the visual itself.
NOTE_06 // SEMANTIC MODELS
2026-10-11·10 min read

Incremental Refresh in Power BI: Setup, Pro Licence Limits and Common Failures

Incremental refresh works on Pro. Here is how to set it up correctly, what you only get with Premium or Fabric, and why most failed setups come down to query folding.

›Incremental refresh is supported on Pro, PPU, Premium and Embedded. Only the real-time DirectQuery partition needs Premium, PPU or Embedded.
›It needs two date/time parameters named exactly RangeStart and RangeEnd, used to filter the table in Power Query.
NOTE_07 // FABRIC & MODERNIZATION
2026-10-11·9 min read

How to Enable Copilot in Power BI: Licences, Capacity and Tenant Settings

What Copilot in Power BI needs, the exact admin settings to switch on, why it often still does not appear, and how to prepare your models so its answers are right.

›Copilot needs a paid capacity: Fabric F2 or larger, Power BI Premium P1 or larger, or a Fabric Copilot capacity. Trial capacities are not supported.
›The tenant setting “Users can use Copilot and other features powered by Azure OpenAI” is enabled by default; admins can scope it to security groups.
NOTE_08 // SEMANTIC MODELS
2026-10-11·11 min read

Power BI Data Modeling Best Practices: A Star Schema Checklist

The modelling rules that prevent wrong totals, slow reports and unmaintainable measures — as a checklist you can run against any semantic model.

›Model as a star schema: narrow fact tables of events and numbers, surrounded by dimension tables that describe them.
›Use one-to-many relationships filtering from dimension to fact in a single direction; treat bidirectional filters and many-to-many as exceptions.
NOTE_09 // FABRIC & MODERNIZATION
2026-10-10·8 min read

Power BI Q&A Is Retiring in December 2026: What to Replace It With

Every Q&A visual stops working in December 2026. Here is how to find where you use it, what Copilot needs, and how to switch without a broken experience.

›Microsoft announced the Q&A deprecation in December 2025; Q&A experiences and Q&A setup are fully retired in December 2026.
›Existing Q&A visuals in reports, dashboards, mobile and embedded scenarios will stop working — this is not a soft deprecation.
NOTE_010 // FABRIC & MODERNIZATION
2026-10-10·9 min read

P1 to F64: A Right-Sizing Guide Before You Migrate

F64 is the same size as P1 — but same size is not the same as right size. How to measure, price and plan the move off Power BI Premium.

›When a P SKU term ends there is a 30-day grace period, throttling from day 31, and rejected operations from day 91.
›F64 matches P1 in size and keeps the free-viewer benefit, which is why it is the default mapping.
NOTE_011 // FABRIC & MODERNIZATION
2026-10-10·8 min read

Direct Lake Fallback to DirectQuery: How to Detect and Prevent It

Your Direct Lake model is quietly running as DirectQuery. Why it happens, how to see it, and how to stop it — including when you should not.

›Only Direct Lake on SQL endpoints can fall back to DirectQuery; Direct Lake on OneLake fails to refresh instead when guardrails are exceeded.
›Fallback keeps queries working but slower, which is why it often goes unnoticed for months.
NOTE_012 // FABRIC & MODERNIZATION
2026-10-10·8 min read

Fabric Capacity Throttling: How to Read the Capacity Metrics App

Smoothing, overage protection, carryforward and the three throttling stages — and what to look at before you buy a bigger SKU.

›Fabric smooths interactive operations over 5–64 minutes and background operations over 24 hours.
›Throttling starts when future smoothed usage exceeds 10 minutes of capacity: first 20-second interactive delays, then interactive rejection after 60 minutes, then rejection of everything after 24 hours.
NOTE_013 // FABRIC & MODERNIZATION
2026-10-10·7 min read

Default Semantic Models Are Gone: Cleaning Up the Orphans

Fabric stopped creating default semantic models and turned the existing ones into ordinary models. Here is how to find them, decide which to keep, and govern the rest.

›Since September 5, 2025, Fabric no longer creates a default semantic model when you create a warehouse, lakehouse or mirrored item.
›By November 30, 2025, existing default semantic models were disconnected from their items and became independent semantic models.
NOTE_014 // DAX & PERFORMANCE
2026-10-10·9 min read

“Visual Has Exceeded the Available Resources”: Causes and Fixes

The error means a query hit a time or memory limit. Here is how to find which one, capture the offending query, and fix the measure or visual behind it.

›The error appears when a visual's query runs too long or uses too much memory; the Power BI Service's default query timeout is 225 seconds.
›Memory per query is capped per capacity — for example 1 GB on F2 and 10 GB on F64 — and Power BI Desktop can simulate limits too.
NOTE_015 // ENGINEERING WORKFLOW
2026-10-10·10 min read

Convert PBIX to PBIP and PBIR: A Team Migration Plan

Saving as PBIP takes a minute. Changing how a team works takes a plan. Repository layout, what to commit, branching, reviews and a phased rollout.

›PBIP saves a report and model as folders of text: TMDL for the semantic model and PBIR for the report, which Microsoft documents as generally available and the default report format.
›PBIX remains the single-file format; PBIP is for content a team maintains together.
NOTE_016 // DAX & PERFORMANCE
2026-10-10·8 min read

ALL vs REMOVEFILTERS: Why Your % of Total Is Wrong

Percent-of-total measures go wrong in three predictable ways. What ALL and REMOVEFILTERS actually do, and which one you want for each kind of total.

›As a CALCULATE modifier, REMOVEFILTERS and ALL do the same thing; REMOVEFILTERS just says so more clearly and cannot be used as a table function.
›ALL on the fact table removes filters from every dimension related to it, so the denominator ignores your year and region slicers.
NOTE_017 // DAX & PERFORMANCE
2026-03-01·6 min read

Why This DAX Measure Is Slow — and How I’d Diagnose It

A step-by-step investigation into context transitions, VertiPaq query plans, and storage engine callback loops.

›Most slow DAX measures are not caused by complex arithmetic, but by the Formula Engine (FE) taking over work that the Storage Engine (SE) should perform.
›Invoking a measure inside row iterators (like SUMX or FILTER) forces an implicit context transition for every single row in the iteration.
NOTE_018 // SEMANTIC MODELS
2026-02-18·8 min read

Anatomy of a Slow Power BI Model: 4 Hidden Flaws Behind an 18-Second Dashboard

The dashboard took 18 seconds to render. The visual wasn't the issue. 4 hidden architectural flaws were.

›A 3.8 GB semantic model was compressed to 420 MB simply by removing 3 unused high-cardinality GUID columns and splitting date-time fields.
›Bidirectional relationships between dimension tables created cross-filtering loops that caused the engine to perform full table scans.
NOTE_019 // ENGINEERING WORKFLOW
2026-01-24·7 min read

Why TMDL Changes the Power BI Development Workflow

Moving away from binary .pbix files to text-based Tabular Model Definition Language, Git branching, and pull requests.

›Monolithic .pbix binary files prevent simultaneous multi-developer collaboration, code reviews, and Git change tracking.
›TMDL breaks the semantic model into human-readable, plain-text files organized by table, measure, and relationship.
NOTE_020 // ARCHITECTURE
2026-01-10·9 min read

7 Common Semantic Model Problems (and How to Fix Them)

From bidirectional relationship cascades to auto date/time table bloat: the architectural anti-patterns hurting enterprise models.

›Snowflake dimension normalization causes needless table joins and slows down VertiPaq decompression.
›Many-to-many relationships through bridge tables introduce ambiguous filter propagation and silent double-counting.

Engineering notes for Power BI builders.

Practical ideas on DAX, semantic models, performance, Fabric, and modern Power BI development.

1–2 useful notes a month. Unsubscribe at any time.