A monthly meeting, three reports, three different revenue figures. Finance has one number, sales has another, and the number on the dashboard matches neither. Everyone assumes somebody is wrong, so the meeting becomes an argument about the data rather than a decision about the business.
Usually nobody is wrong. Each department is measuring something slightly different, and each is right within its own definition.
The same word, three meanings
Take revenue. Finance usually means invoiced and recognised in the period, excluding VAT, net of credit notes. Sales usually means the value of deals marked won, on the date they were marked, including everything expected to invoice later. An e-commerce platform usually means orders placed, including VAT, including the ones later refunded.
Three honest answers to what looks like one question. The gap between them is not an error to find; it is a definition to agree.
The same applies to almost every measure a business reports on:
- A lead: anyone who filled in a form, or only those who passed qualification?
- A new customer: first order placed, or first order paid?
- Churn: cancelled, lapsed after a period of inactivity, or failed to renew?
- Active user: logged in once this month, or did something meaningful?
- Cost per acquisition: ad spend only, or including the retainer and the tools?
Why a dashboard does not fix this
The instinct when reports disagree is to buy reporting software. It is the wrong order. A dashboard built on top of unresolved definitions does not settle the argument — it publishes the argument faster, with better typography, and gives it the authority of a chart.
Worse, it usually adds a fourth number. Now there are three departmental figures and a dashboard figure, and the dashboard becomes the thing nobody trusts.
What a metric needs before it goes on a report
Four things, written down, for every measure that anyone makes a decision on:
- An owner — one named person who decides what it means when there is a question.
- A written definition in plain language, including the date it is counted on.
- A source system, named. If two systems hold it, one of them is the authority and the other is a copy.
- An exclusion list. Internal orders, test transactions, refunds, staff accounts, the one enormous deal that distorts every average.
Do you actually need a data warehouse?
Not always. A data warehouse earns its place in two situations, and plenty of businesses are in neither yet.
The first is when answering a single question requires joining two systems. What a customer acquired through a particular campaign is worth after twelve months cannot be answered by the ad platform or the accounting system alone. Neither can see the other, so something has to hold both, which is what a growth intelligence layer is for.
The second is when manual assembly has become the bottleneck. If reports take days to produce, they are days old before anyone reads them, and the business is steering using a rear-view mirror.
If neither is true, better definitions and a tidier spreadsheet will serve you for another year, and we will say so.
The spreadsheet is telling you something
Most businesses have one: the workbook that holds the real figures, maintained by one person, consulted before anything important. It is easy to treat it as a problem to eliminate. It is more useful to treat it as evidence.
A spreadsheet survives because the systems are not trusted, or because they cannot answer the question, or because the definition that matters exists only in one person’s head. Those are three different problems with three different fixes, and the spreadsheet is the clue to which one you have.
It also carries a risk worth naming: the person maintaining it is the integration. It has no error handling, no audit trail, and it goes on leave.
A sensible order of work
- Agree definitions for the ten or so measures leadership actually acts on.
- Name the source of record for each, and write down the exclusions.
- Reconcile each definition against the system it comes from, once, by hand, which is also where an analytics review earns its keep. Expect surprises.
- Only then decide whether the joins you need justify a warehouse, or whether scheduled exports will do.
- Build reporting against the definitions, not against whatever the tools make easy.
Done in that order, the reporting tool becomes a detail. Done in the opposite order, the reporting tool becomes the fourth opinion in the room.
Frequently asked questions
How long does agreeing metric definitions take?
For most businesses it is a few focused sessions rather than a project. The time goes into reconciling each definition against its source system, which is where the disagreements surface and get settled.
Can we use our existing tools instead of a warehouse?
Often, yes. If every question you need answered sits inside one system, scheduled exports and a well-defined spreadsheet are enough. A warehouse earns its cost when questions require joining systems that cannot see each other.
Who should own a metric definition?
One named person, usually whoever is accountable for the outcome it measures. Shared ownership means the question reopens every time the number is questioned.
