The tolerated pain
The Monday leadership meeting opens with a familiar argument. Sales reports 212 new customers for the quarter. Finance reports 178. Operations says 190 accounts were activated. The marketing dashboard says 240 sign-ups. All four are accurate against their own rules, and all four are presented with confidence. The first twenty minutes go to establishing whose number is "right". The decision the meeting was called to make, whether to hire another account manager, is postponed to next week.
This is not a data-quality accident. It is what happens when each department answers the same question with its own definition, its own source system and its own cut-off date, and nobody has agreed which one the business uses. The related problem of untrusted numbers is covered in untrusted business numbers and the cost of unreliable decisions. This article is about the specific habit of arguing over the figure instead of acting on it, and the small amount of agreement that removes it.
Why departments report different numbers
Different numbers for the same metric almost always come from one of five sources.
- Different definitions. A "customer" might be someone who has enquired, someone who has signed, someone who has paid or someone who has been active in the last ninety days.
- Different source systems. Sales counts from the CRM, finance from the accounting package, operations from the delivery system. Each is correct for its own purpose.
- Different timing. One report counts at the order date, another at invoice date, another at payment date. A month-end cut-off on the 25th versus the 31st will never reconcile.
- Different filters and adjustments. Credit notes, cancelled orders, internal accounts and test records are included in one report and removed from another.
- Manual steps. A figure copied into a spreadsheet and adjusted "just this once" becomes a version nobody can trace. See why executives get different answers to the same KPI question.
None of these is dishonest. They are the natural product of teams building the reports they need without a shared dictionary.
Why businesses tolerate it
- Everyone has a plausible explanation. "Finance counts differently" sounds like a reason, so the discrepancy is accepted.
- Definitions live in people's heads. The rule exists, but it is in the analyst's memory or a slide deck. Compare metric definitions that live in slide decks, not governed models.
- Agreeing a definition feels political. Choosing one version means one department's number goes down.
- The cost is spread thinly. Twenty minutes in a meeting looks small. Multiplied across a year and across the leadership team, it is not.
- Dashboards are easy to build. More dashboards without a shared definition simply produce more versions.
Operational symptoms
- meetings that start with reconciling figures instead of discussing decisions;
- managers bringing their own spreadsheets "because the official report is wrong";
- analysts spending the first days of each month explaining differences;
- different numbers in the board pack, the management accounts and the sales report;
- decisions postponed pending "a clean number";
- staff who stop trusting the dashboards and fall back on instinct.
How the pain becomes money
- Meeting time: senior people, whose time is the most expensive in the business, spend it comparing figures.
- Reconciliation labour: analysts and finance staff rebuild and explain numbers every month; see rebuilding recurring reports by hand.
- Decision delay: a hiring, pricing or stock decision that waits a week or a month has a cost in lost sales, extra stock or missed supplier terms.
- Wrong decisions: when one department's version drives the outcome, the choice may be based on a number the rest of the business does not recognise.
- Audit and lender cost: if a figure cannot be traced from the report to its source, external reviewers spend longer and charge more. The problem is explored in untraceable reporting numbers and the audit cost.
An illustrative cost model
This example is illustrative. All figures are assumptions for a hypothetical 80-person distribution business in Gauteng.
| Assumption | Value |
|---|---|
| Weekly leadership meeting attendees | 7 |
| Minutes per meeting spent on number disputes | 25 |
| Average loaded hourly cost of attendee | R650 |
| Meetings per year | 46 |
| Analyst and finance days per month spent reconciling | 4 days (8 hours each) at R280 an hour |
| Decisions delayed per year by one week or more | 6, each costing roughly R15,000 in lost margin or extra cost |
- Meeting time: 7 × (25 ÷ 60) × R650 × 46 = about R87,200
- Reconciliation: 4 × 8 × R280 × 12 = R107,520
- Delayed decisions: 6 × R15,000 = R90,000
- Illustrative total: about R285,000 a year
The decision-delay line is the softest. You can firm it up by listing the decisions that were postponed last year and estimating what each week of delay cost.
When it becomes commercially urgent
- the business is preparing for a funding round, sale or lender review;
- a board or investor asks for a monthly KPI pack;
- incentives or commissions are tied to a figure that more than one department calculates;
- the business introduces AI or automated reporting, which will repeat whichever definition it is given;
- a major decision, such as a new site or product line, depends on a number leadership does not agree on.
What better looks like
The fix is smaller than most owners expect. It is mostly agreement, with some technology to hold the agreement in place.
- Pick the ten numbers that matter. Revenue, gross margin, active customers, order backlog, on-time delivery and cash are typical. Leave the rest.
- Write one definition for each. Include the source system, the date used, what is included and excluded, and who owns the definition. One page is enough.
- Name an owner. The person who owns a metric answers questions about it and approves any change.
- Calculate it once. Put the logic in one place that every report reads from. In Microsoft's Power BI, for example, a measure holds a calculation that reports can reuse, and a shared semantic model lets many reports draw on one set of data and logic. The same principle applies in any tool, including a well-controlled spreadsheet.
- Show the definition next to the number. A small footnote prevents most arguments.
- Keep the legitimate variants. Finance may still need cash-basis revenue. Name it differently, such as "invoiced revenue" and "cash received", rather than forcing one number to do two jobs.
- Review quarterly. Definitions change when the business changes, and the change should be recorded, not silent.
The Insights article why your reports cannot agree on the same number goes further into the causes.
Practical next actions
- Choose one metric that has caused an argument recently and write down how each department calculates it.
- Meet for an hour with the owners of those versions to agree one definition or two clearly named ones.
- Record the definition, source and owner in a shared one-page dictionary.
- Change the main report to use the agreed definition and label it.
- Retire the old versions from circulation and tell people.
- Repeat for the next metric each month until the key ten are covered.
Bottom line
When departments report different numbers, the business pays twice: once to produce the competing versions and again to argue about them. The number itself is rarely the real problem. The missing piece is a shared definition with an owner, held in one place. Agreeing that for a handful of important metrics moves meetings from "whose figure is right?" to "what do we do about it?", which is the point of having the figures at all.
Sources and further reading
- Measures in Power BI Desktop - Power BI (last verified 10 October 2026)
- Semantic Models in the Power BI Service - Power BI (last verified 10 October 2026)
- Power BI usage scenarios: Enterprise BI - Power BI (last verified 10 October 2026)
Related Mellorca reading
- Four Tools, One Job: The Real Cost of Departments Buying Their Own Software
- Sales, Finance and Operations Each With Their Own Customer: The Cost
- Two Departments, Same Work: The Payroll Cost of Duplicated Effort