Family · 5 min read
FP&A variance analysis: four gaps a revenue bridge never shows
A revenue bridge explains the top line. Below it sit four other gaps that decide whether the year lands: what you spent, what you paid your people, what you invested, and what your balance sheet did to your cash. None of them is in a sales export, and each one needs a file of its own.
Four questions a plan cannot answer alone
Most planning cycles produce one number per line and stop there. Spending is 2,200 over. Payroll is 20,900 under. The variance report says so, and the meeting spends forty minutes guessing why.
The four analyses in this family do the part that comes before the guessing: they take the gap apart along the only lines the file can actually support, and they say which declaration each split rests on. That last point is what makes the answer defendable three weeks later, when nobody remembers what was assumed.
The four analyses
OPEX variance compares committed spending against a plan, account by account, at a comparison level you choose. On the sample ledger, budget 30,500 against actual 32,700 - a gap of 2,200 - and the level you pick decides whether a relocated cost reads as one line or as two opposite ones.
Headcount and payroll splits a payroll gap into average cost per person and headcount, plus positions filled without a plan and planned positions left unfilled. Four effects that add up to the gap exactly.
CAPEX and depreciation answers two questions that are constantly confused: what was committed against what was planned, and what the resulting charge is month by month. A project perfectly on budget changes the second completely if it goes into service a month later.
Working capital and cash turns movements in receivables, inventory and payables into their cash effect. On the sample file, working capital rose from 102,000 to 109,000 - which consumed 7,000 of cash.
What you declare, and why
Each of the four asks for something the file cannot supply, and refuses to guess it. That is not caution for its own sake: every one of these is a decision where a wrong default produces a plausible number that no total would contradict.
- Which scenario is the plan - every one of the four asks. A file that writes FY26 and FY26 R2 does not say which one happened.
- How your file writes costs - read the wrong way round, every overrun becomes a saving.
- What each account type means - a supplier debt filed among the assets would consume cash instead of releasing it.
- When the first depreciation charge falls - the month of service, or the month after. The gap between the two conventions is a whole monthly charge on every project.
- That one period is one month - a useful life is counted in months, and quarters would divide every charge by three.
Each declaration travels with the result: into the screen, into the workbook, and into the review pack. A figure whose convention is not written next to it cannot be checked by the person reading it.
How this differs from the commercial bridges
A price, volume and mix bridge and the pricing and margin family both explain what happened to revenue and margin, from a sales export. This family never touches a sales file. It reads ledgers, payroll extracts, investment registers and balances - four contracts that have nothing in common with a product line.
The nearest neighbour is actual vs budget, and the difference is worth stating: that one compares revenue against a plan and splits it into price, volume and mix when the file allows. These four compare costs, people, investment and balances, and split them along entirely different lines. Reading both is normal; confusing them is not.
Build yours
Read your own four
Bring each file into a monthly review and read them next to your commercial analyses. It all runs inside your browser: nothing is uploaded, which you can check in the Network panel while you work.
Start a monthly reviewWhat none of them will do
None of them names a cause. A ledger records what was spent, not why. If your file carries a column that comments a line, it is shown exactly as written, under a heading that concludes nothing.
None of them recommends anything. No headcount target, no spending cut, no investment priority. A payroll file does not say whether someone is well paid.
None of them forecasts. They read points that already exist. Nothing is extrapolated to a period the file does not contain.
None of them produces a cash flow statement. The working capital analysis explains the cash impact of working capital, and stops there: profit, investment, financing, tax and currency are outside it.
None of them touches tax, impairment, currency or lease accounting. Depreciation is straight line with no residual value. These are stated boundaries, not gaps waiting to be filled.
Questions
Can one file feed all four?
No, and that is deliberate. The four contracts share only a period and a scenario label. Each analysis reads its own file, and the product never joins two of them.
Which one should I run first?
OPEX variance, usually. It needs the fewest columns - period, scenario, account, amount - and it is the one most often already exported from the ledger.
Does my file leave the browser?
No. The whole calculation runs in the page, on your machine. There is no upload, no account, and no server that could hold your figures. How to verify that in two minutes.