Every business runs on a quiet argument between three numbers: what you planned to spend and earn, what you now think you will spend and earn, and what you actually did. Financial planning and analysis, FP&A, is the function that keeps that argument honest, and variance analysis is how it settles. Get it right and the numbers become a steering wheel. Get it wrong and they become a rear-view mirror you stare at while the road bends.

The quick version

  • FP&A is the team and the process that build the budget, maintain the forecast, and explain the gap between plan and reality so leaders can decide what to do next.
  • A budget is a fixed annual plan and a commitment; a forecast is a living best-guess of where you will actually land, updated as facts change. They answer different questions and should not be confused.
  • Variance analysis compares actuals to plan and splits the difference into causes, sold more units, paid more per unit, spent on the wrong things, so you act on reasons, not totals.
  • The trap is treating the forecast as a test to pass. People game numbers they are judged against, so a forecast loaded with optimism or padding tells you about incentives, not the future.

The idea in depth: FP&A is decision support, not bookkeeping

It helps to separate two jobs that both involve numbers. Accounting records what happened, it looks backward and has to be exactly right. FP&A looks forward and has to be usefully approximate: it takes the historical record, adds assumptions about the future, and produces a plan and a running prediction that managers steer by. (For the deeper split, see management vs financial accounting.) Confusing the two is a common and expensive mistake, demanding accounting-grade precision from a forecast wastes weeks and still gets you a number that is wrong, because the future is genuinely uncertain.

So judge a forecast by whether it changes a decision, not by how many decimal places it carries. If two versions of the revenue forecast lead to the same hiring plan, the difference between them is noise. Ask of every number on the page: what would we do differently if this were 10% higher or lower? If the answer is "nothing," stop polishing it.

Budget versus forecast: a commitment versus a prediction

The most useful distinction in the whole discipline is also the most often blurred. A budget is set once a year, ahead of time, and functions as a target and a resource allocation, it is a promise about how money will be spent. A forecast is the current honest estimate of where the year will actually finish, refreshed as new facts arrive. Both matter, but for different reasons: the budget holds people accountable; the forecast keeps them informed.

The sharpest critique of leaning too hard on the annual budget came from Jeremy Hope and Robin Fraser, who with Peter Bunce founded the Beyond Budgeting Round Table in 1998. In Beyond Budgeting (2003) they argued that the fixed annual budget is slow, centralising, and quietly corrupting: when one number is at once the target, the forecast, and the basis for bonuses, people stop telling you the truth about any of it. Their alternative leans on rolling forecasts, a forecast that always looks four to six quarters ahead and is re-cut each quarter, so planning never falls off a cliff at year-end.

The practical fix is to separate the jobs the budget is doing: keep a target if you need one, but maintain a forecast allowed to disagree with it, and never let the number someone is rewarded for hitting double as the number you plan with. The honest limitation: "beyond budgeting" is a philosophy with persuasive case studies, not a settled empirical law, and later work (a 2024 case study in European Accounting Review) found that in practice rolling forecasts tend to complement budgets across time horizons rather than abolish them. Lean that way; don't scrap your budget on Monday.

flowchart LR
  A(["Budget
set once, a year ahead
= target & commitment"]) --> C{"They disagree.
Why?"}
  B(["Forecast
updated each quarter
= honest prediction"]) --> C
  C --> D(["Variance analysis
splits the gap into causes"])
  D --> E(["A decision:
act, replan, or ignore as noise"])
The three numbers and the loop between them, the budget commits, the forecast predicts, variance analysis explains the gap. Leaders Loop

Variance analysis: turning a gap into a reason

A variance is simply the difference between what you planned and what happened. The skill is decomposition, splitting one big number into the smaller causes that produced it, so you act on the right one. The classic technique comes from standard costing and the flexible budget, both staples of management-accounting texts such as Horngren's Cost Accounting. A standard cost is a predetermined expected cost per unit; a flexible budget is the budget recalculated at the volume you actually achieved, rather than the volume you planned.

That recalculation is the move that turns noise into signal. Suppose you budgeted to make 1,000 units and spend a certain amount on materials, but demand surged and you made 1,300. Of course you spent more on materials, you made more units. Comparing actual spend to the original 1,000-unit budget tells you almost nothing. The flexible budget asks instead: what should 1,300 units have cost? The gap between actuals and that flexed figure is the part that is genuinely about efficiency or price, not volume. As AccountingTools puts it, the flexible-budget variance isolates performance from the effect of activity levels.

So never report a raw budget-versus-actual variance without first asking whether volume changed. Split every meaningful variance into a volume part (we did more or less) and a rate part (each thing cost more or less than expected), and chase only the part you can actually control. A favourable cost variance caused by under-producing is not a win; a small unfavourable variance hiding a large efficiency loss masked by extra volume is not the non-event it looks like.

A variance you cannot explain is not a number, it is a question you have not asked yet.

The honest limitation: variance analysis tells you what differed, never why. The decomposition is arithmetic; the diagnosis is judgement. A price variance might be a procurement failure, a supplier renegotiation, or a deliberate decision to pay more for faster delivery, the maths cannot tell them apart. And there is a subtler trap, the one Hope and Fraser flagged: if managers know an unfavourable variance triggers an awkward meeting, they learn to forecast conservatively so reality always beats the plan. The technique is only as honest as the culture it sits inside.

The bias problem: your forecast is also a mirror of your incentives

Two predictable distortions show up in almost every forecasting process, pulling in opposite directions. Optimism bias is the systematic over-forecast, common where salespeople or founders project confidence into the pipeline. Sandbagging is the deliberate under-forecast, common where bonuses reward beating the plan, so people quietly set one that is easy to beat. Both are forecast bias: a consistent, directional error rather than random scatter. The practitioner literature (for example, Finance Alliance) is blunt that both wreck planning, optimism over-commits resources, sandbagging starves growth.

So measure your own bias before you trust your own forecast. Keep a simple log: each period, record what you forecast and what happened, then look at the average error over time. Random noise averages toward zero; a persistent positive or negative average is bias, and it is correctable. If your team has over-forecast revenue by roughly the same margin four quarters running, adjust the fifth forecast down on the evidence, not because anyone is dishonest, but because the pattern is data. Naming the bias out loud, before the number is locked, is most of the cure.

A worked example

Take a small coffee-equipment company, call it Latitude. (Illustrative figures throughout; this is a teaching example, not real accounts.) For the quarter, FP&A budgeted to sell 1,000 machines at a standard materials cost of £60 each, a budgeted materials spend of £60,000. The quarter closes: Latitude actually sold 1,300 machines and spent £83,200 on materials. The raw variance is £23,200 over budget. A nervous manager calls a cost-control meeting.

The meeting is premature, because the raw number mixes two unrelated stories. First, flex the budget to actual volume: 1,300 machines × £60 standard = £78,000. That £18,000 of the overspend is simply the cost of making 300 extra machines you were thrilled to sell, it is a volume variance, and it is good news. The part that deserves attention is the rest: actual £83,200 versus the flexed £78,000, a £5,200 unfavourable flexible-budget variance. That £5,200 is the real question, roughly £4 of extra materials cost per machine. Was it a supplier price rise, scrap from a faulty batch, or a deliberate switch to a better component? The arithmetic has pointed the finger; now a human investigates the one number worth investigating.

flowchart TD
  A(["Raw variance: £23,200 over budget
(actual £83,200 vs budget £60,000)"]) --> B{"Did volume change?
1,000 planned → 1,300 sold"}
  B --> C(["Flex the budget:
1,300 × £60 = £78,000"])
  C --> D(["Volume variance: £18,000
cost of selling 300 more, good news"])
  C --> E(["Flexible-budget variance: £5,200
~£4/machine, the real question"])
  E --> F(["Investigate just this:
price rise? scrap? better part?"])
One scary total, split into a volume part you can ignore and a rate part worth a meeting. Leaders Loop

The lesson is the order of operations: flex first, decompose second, diagnose third. Reverse it and Latitude burns an afternoon "controlling costs" on £18,000 of spend that was the direct result of a great quarter, while the genuine £5,200 efficiency signal goes unexamined.

Frequently asked questions

What does an FP&A team actually do day to day?

Three recurring jobs. They build the annual budget with the rest of the business; they maintain a forecast that updates the year's expected landing point as actuals come in; and they run the monthly close-and-explain, comparing actuals to plan, decomposing the variances, and telling leaders what the gaps mean and what to do about them. The fourth, less visible job is decision support: modelling specific choices, such as whether to hire, how a price change flows through, or whether a project clears its hurdle.

What's the difference between a forecast and a budget, in one line?

The budget is what you committed to before the year started and rarely changes; the forecast is your best current estimate of where you will actually finish and changes whenever the facts do. The budget is for accountability; the forecast is for steering.

What is a rolling forecast and is it worth the effort?

A rolling forecast always looks a fixed distance ahead, commonly four to six quarters, and is re-cut each period, so you never run out of forward visibility at year-end the way a static annual plan does. It is more work, and it earns its keep most in fast-moving or seasonal businesses where the annual budget is stale by March. In a stable business, a lighter quarterly reforecast may be enough.

Why split a variance into volume and rate?

Because the two have completely different meanings and owners. A volume variance reflects how much you sold or made, often outside the cost team's control and frequently good news. A rate (or price/efficiency) variance reflects what each unit cost versus the standard, that is the part that signals a real operational issue. Lumping them together produces either false alarms or false comfort.

How do I stop people gaming the forecast?

Separate the numbers that serve different masters. Don't make the forecast you plan with the same number someone is bonused on, because then it stops being a prediction and becomes a negotiation. Measure forecast bias openly over time so a pattern of optimism or sandbagging is visible and discussable rather than personal, and reward forecast accuracy, not just hitting a low bar.

Related in the Toolkit

Forecasting sits on top of how you read the business's actual numbers, the statements that feed every forecast (financial statements), and shades into the discipline that sets the plan it measures against (budgeting).

Where to go next