Get the numbers to agree.
We trace the disputed measures to source, settle the definitions with the people who argue about them, and leave you with a model the reporting can be rebuilt on.
3–4 weeks · scoped to the reports a decision depends on
Two reports, one number, no agreement.
The cause is almost never the reporting tool. It is that nobody agreed what a customer is, or when revenue counts, or whether cancelled orders come out of last month.
- 01Which measures are actually disputed, and who is arguing — the list is usually shorter than it feels.
- 02Where each figure comes from, traced to the system that produces it rather than the report that shows it.
- 03Whether a definition exists in writing, and whether anyone owns it.
- 04Timing: invoice date against dispatch, accrual against cash, and where the period boundaries sit.
- 05Whether a semantic model exists, or every workbook re-implements the logic.
- 06How many versions of the same report are in circulation, and which one leadership actually opens.
- 07Refresh and freshness: what the numbers promise and what they deliver.
- 08What breaks at month-end, and how much of the close is spent reconciling rather than closing.
What lands on your desk
A definitions register
The measures that matter, written down, agreed across the business, with a name against each one.
The reconciliation
The tie-out between sources for the disputed measures, and the reason each gap exists — which usually ends the argument on its own.
A semantic model design
One governed model where a measure is defined once, so reports inherit it rather than reproducing your disagreements faster.
A rebuild plan
Which reports are rebuilt, which are retired, and which are left alone, with effort against each.
Pipeline checks
The validation and reconciliation checks to put in place so the next gap surfaces before a board meeting rather than during one.
A self-service boundary
What analysts should build on, what is certified, and where the line sits — so you do not end up with a fourth version of revenue.
Four weeks to numbers people accept.
- Week one
Scope to the decisions the numbers serve, gather the reports in circulation, and identify what is genuinely in dispute.
- Week two
Trace each disputed figure to source and establish whether the difference is a definition, a timing rule or a data fault.
- Week three
Work the definitions through with finance and operations together, because settling it in one room is faster than three rounds of email.
- Week four
The model design, the rebuild plan and the checks — with the register handed over in a form your team maintains.
Data integration and analytics hub — commodity trading and risk. Data had sprawled across systems and what analysis was possible ran on figures that were already out of date. We built the integration layer and the analytics capability on top of it, implemented on Posit and connected to inventory, pricing and related sources.
Before you ask
Do we need a new reporting platform?
Usually not, and we will say so. Most of this is definition and modelling work that lands on whatever you already own.
Our finance team says their number is right.
They are often both right, which is the point. Two correct answers to two different questions is the most common finding, and it is settled by agreeing which question the report answers.
Bring us the two reports.
Forty-five minutes on which numbers are disputed and what decision is waiting on them.