Power BI portfolio — Nordic Boardroom reports that survive a cold open

The question

When someone opens a report cold on a Monday, with two minutes and no walkthrough, what has to work: the theme, or the path from "are we okay?" to "who do we act on?"

I spent years in Nordic banking on PD models, scorecards, and expected-credit-loss style work. The habit that survived was not a color palette. It was: take a messy commercial question, land it in a shape that repeats, and write the caveats down instead of waving them away. This portfolio is that habit practiced in public — churn, sales health, bank engagement, clinical risk — under one Nordic Boardroom grammar.

See it: Portfolio → Power BI
Repo: github.com/AlexTouvras/powerbi-portfolio

Gold before chrome

In data-platform talk, bronze sits close to the extract, silver is cleaned and joined, and gold is the curated layer you would trust for a KPI — usually fact and dimension tables shaped so measures stay honest.

When I say gold here, I mean that layer, not a brand. If the grain is wrong, page polish only makes the lie prettier. So every report starts with gold tables and measures, then a semantic model (relationships, DAX, what a slicer is allowed to filter), then pages. Screenshots come last.

I export each page to PNG so Orbit can show the work without embedding a live Fabric workspace for every visitor. npm run powerbi:sync (and the GitHub Action) refreshes those images plus metadata into this site. If the PNG drifts from what is committed, the site becomes a prettier lie than the repo.

The page path I refuse to invent twice

Landing states the question and who the report is for — a poster cover, not a matrix. Pulse answers "are we okay?" with a few durable KPIs. Drivers answers "why?" (influencers, decomposition, mix). Queue answers "who do we call next?" Context holds model limits: storage mode, what a score means, what the numbers are not.

That order is deliberate. A dense matrix on page one loses people who only have two minutes. A pretty Pulse with no queue loses people who need a next action. Same path every domain means a new report reuses the grammar instead of inventing navigation.

I don't pretend Import mode is Direct Lake, or that a propensity score (churn, readmit, and the rest) is ground truth. Context is where that honesty lives.

What scales is the kit, not one hero report

Theme, page grammar, and export rules live under _shared. Each domain report plugs into that kit. New work should start from shared patterns, not a blank canvas and a one-off layout.

Churn is the path under load: risk on Pulse, Key Influencers and decomposition on Drivers, an at-risk queue sorted by score, then Context. Sales leans status-first, then product and concentration. Bank and care reports reuse the same reader path with their own gold. Nordic Equity also has a clickable TradingView-style board at heatmap-web-five.vercel.app — logos, day-change %, sector zoom — because native Power BI treemaps don't carry that chrome well.

What still fights me

I still argue with myself about Direct Lake vs Import for demos. Import keeps the cold-open story simple here: the PNG is what shipped, and the model is sized for a laptop walkthrough, not a capacity theater.

I also care more about boring correctness than polish: measures that match the grain of the decision, slicers that don't blow up the model, a Context page that says what the model can't do, screenshots that track the committed definition.

Full diagram set: Power BI portfolio — architecture.

Takeaway

A report earns its slot when Landing → Pulse → Drivers → Queue works without a narrator. Theme is the easy part. The hard part is the contract between gold tables, measures, and the page that names who to contact next.