Project · finance business partnering

Business Unit Performance Dashboard

An illustrative management-reporting cockpit for a generic Nordic wholesale-bank division with five product lines. Three layers of the monthly pack — KPI tiles, FTP-aware NII variance waterfall, per-line performance table — plus an interactive sensitivity panel and a ROAC-vs-hurdle view, all reading from a synthetic dataset built in Python and queried through SQLite.

Illustrative synthetic data. Not affiliated with or representative of any real bank. All numbers generated by pipeline/generate_data.py.

Headline KPIs

—

NII variance — YTD Actual vs Budget

Decomposition: volume · customer margin · FTP · other

FTP-aware NII bridge

Budget NII → effects → Actual NII (NOK millions)

Three things leadership should know

Auto-generated from the YTD variance data

Monthly trend — last 24 months

Solid lines: actuals through Apr 2026. Dashed: forecast through Dec 2026.

Per product line — YTD performance

Actual vs Budget · annualised ROAC

Product line	Income	Opex	Impairment	PBT	C/I	Allocated capital	ROAC

ROAC vs hurdle rate

Hurdle = 11.0% · annualised YTD return on allocated capital

Sensitivity — interest-bearing book

Move volume growth, customer spread, and FTP rate to see live NII and division ROAC impact

Lending volume growth 0.0%

vs YTD Actual baseline · applies to LSF + TB books

Customer spread shift 0 bps

parallel shift to customer rates on lending and deposits

FTP funding-cost shift 0 bps

parallel shift in the internal transfer rate

NII (YTD, NOK m)

—

PBT (YTD, NOK m)

—

Division ROAC (annualised)

—

vs hurdle (11.0%)

—

How it's built

Pipeline lives at prosjekter/forretningsenhet-dashboard/pipeline/

A pandas script generates 24 months of synthetic Actual / Budget / Forecast figures and writes them to a single fact table in SQLite. The reporting blocks above are produced by readable .sql files (one per view), and a small Python step rolls the results into the data.json this page consumes. Commentary is rule-based; an optional Anthropic-API variant is wired in behind a flag and off by default so the page works without any key.

Python · pandas SQLite SQL Chart.js Vanilla JS

← Back to projects