Services
Three engagements, priced from a floor rather than by the hour. Scope is agreed in writing before work starts, and the deliverables below are what you actually receive — not a category of activity.
Data audit
From CHF 6’000 · fixed, two weeks
Two weeks to find out what your numbers are actually worth, and what it would take to trust them.
Most reporting problems are not reporting problems. They are definition problems, ownership problems, or a spreadsheet somebody built in 2019 that nobody dares to touch. Until you know which one you have, buying software is guessing.
The audit is deliberately short and fixed in price. Two weeks, a fixed fee, and a document at the end that you own outright — whether or not you ever work with me again.
How it runs
I start from the report your leadership team actually looks at, and work backwards. Every figure on it gets traced to the system it came from. Where two systems disagree, I find out why, and write down which one is right and what it would take to make that permanent.
You get an hour of my time each week for questions, and a written summary at the end. No slideware.
What it does not include
The audit does not fix anything. It tells you precisely what is broken and what fixing it costs. Some clients take that document and fix things themselves with their own team; that is a perfectly good outcome, and the document is written so it is possible.
What you receive
- A written inventory of every source your reporting touches, and who owns it
- A list of the specific defects — duplicates, gaps, silent format changes, contradictory definitions
- One reconciled figure, traced end to end, so you can see where the discrepancy comes from
- A costed plan: what to fix, in what order, and what each step buys you
Pipeline build
From CHF 18’000 · from, typical engagement six to ten weeks
The machinery that turns scattered sources into one validated dataset, with the checks that tell you when it breaks.
A dashboard is only as honest as the pipeline under it. This is that pipeline: the scheduled, tested, boring machinery that takes whatever your source systems emit and produces one dataset your finance and operations people both agree on.
The part most people skip
Failure handling. A pipeline that silently serves stale numbers is worse than no pipeline, because people make decisions on it. Every build I ship asserts on the data’s own timestamps and stops rather than publishing a stale snapshot. If the feed dies at 03:00, you find out at 08:00 from an alert, not in November from an auditor.
Built to be handed over
I write these in Python, with tests, in your repository. No proprietary runtime, no per-seat licence, nothing that stops working when the engagement ends. The handover session is part of the price, not an extra.
If your team has no one to hand it to, say so early — that changes the design, and it is better to know at the start than at the end.
What you receive
- Automated ingestion from each source, scheduled and logged
- Validation rules that reject bad records loudly instead of passing them through quietly
- A freshness check that fails when the data stops arriving, rather than serving yesterday as though it were today
- Tests covering the ugly cases: wrong encodings, missing columns, duplicate keys, empty dates
- Written handover documentation, so your team can run and change it without me
Executive reporting
From CHF 12’000 · from, typical engagement four to six weeks
One report your leadership team reads without arguing about the numbers first.
The failure mode I am hired to fix is not a missing chart. It is a meeting that spends its first twenty minutes disputing whose figure is correct.
This engagement produces one report, built on top of a pipeline you trust, designed around the specific decisions your leadership actually makes. Not a wall of every metric that could be plotted — the handful that change what somebody does on Monday.
Design is not decoration here
Colour carries meaning or it stays out. Every categorical palette I ship is validated for colourblind separation before it goes anywhere near a page, because roughly one man in twelve cannot reliably distinguish the red and green that most dashboards lean on. Every chart carries a text alternative, so the report survives being read aloud, pasted into an email, or opened by someone using a screen reader.
What I will push back on
Vanity metrics, dual-axis charts, and any number nobody can name a decision for. If a figure survives the question “what would you do differently if this doubled?”, it belongs on the report. If it does not, it belongs in an appendix.
What you receive
- A single report or dashboard, designed around the decisions it supports
- Every figure traceable to its source, with the definition written down
- Accessible by construction: readable on a phone, legible to colourblind readers, usable without a mouse
- A short written guide to reading it, aimed at the people who will use it
Not sure which one?
Start with the call. If the audit is not the right first step, I will say so — including when the honest answer is that your team can fix this without hiring anyone.