What is variance analysis?

Every month, the same drumbeat. The actuals land, and someone has to explain the gap. Why is revenue light? Why did margin slip when volume was up? Why is the cost line over plan when headcount is under? That explanation has a name (variance analysis) and it is, quietly, the task FP&A performs more than any other. It’s also the one most teams stop halfway through: they report the number and never reach the reason.

This is the foundational piece. We’ll define variance analysis, settle the convention that trips everyone up (Actual minus Budget, or Budget minus Actual?) and show what it means to decompose a gap rather than just announce it. The deeper machinery (the step-by-step price–volume–mix split, the waterfall, the leap from variance to cause) gets its own articles. This one is the map.

What variance analysis actually is

Variance analysis is the systematic comparison of actual results against a reference point, and the decomposition of the resulting gap into the underlying drivers that caused it.

Two words in that sentence do the heavy lifting.

Comparison. You cannot have a variance without a baseline. And the baseline you choose changes the question you’re asking, more on that in a moment.

Decomposition. A single number (“revenue is two million light”) is a symptom, not an analysis. The work is breaking that two million into the levers a business can actually pull: price, volume, mix, timing.

A variance you can’t decompose is just a complaint with a currency symbol in front of it.

That second word is what separates analysis from bookkeeping. Anyone can subtract two numbers. The craft is in explaining the difference.

The baseline decides the question

The same actual figure tells you completely different things depending on what you hold it against. Pick the wrong comparison and you’ll answer a question nobody asked.

Mature finance functions don’t pick one. They run several, because each isolates a different failure. A miss against Budget but a hit against Forecast means your plan was wrong, not your execution. That single distinction can change who’s accountable for a number. Which comparison answers which question deserves a closer look of its own. We’ll come back to it.

Actual − Budget, or Budget − Actual?

This is the question that causes more reporting confusion than any other single thing in the discipline, so let’s answer it plainly.

The arithmetic sign is ambiguous. The favourability is not.

Start with the intuitive default, Actual − Budget. Revenue of 52 against a plan of 50 gives +2, and your instinct says good. But apply the same subtraction to a cost line: actual spend of 12 against a plan of 10 also gives +2, and now positive is bad. Same operation, opposite meaning, depending on which row you’re standing on.

That’s why Budget − Actual is common in cost-accounting traditions: for an expense, it flips the sign so that under budget reads as positive. Spend 10 against a plan of 12 and you get −2, an overspend, shown as a negative. Tidy, for costs. Confusing if you apply it to revenue without thinking.

So the raw sign flips meaning depending on the account, which is exactly why professionals don’t trust the sign on its own. They report in terms of Favourable (F) and Unfavourable (U):

A variance is Favourable if it increases profit (more revenue, or less cost) and Unfavourable if it decreases profit. Regardless of the arithmetic sign.

The practical rule that follows: present variances so that favourable is always positive, on every line. That usually means Actual − Budget on revenue, Budget − Actual on costs, and an F/U label on every row so the reader never has to do sign arithmetic in their head. Pick the convention, apply it without exception, and put the label on the page. The goal is a report where positive always means good, because a reader who has to remember “wait, is up good here?” is a reader who will eventually get it wrong.

Decomposition: where the analysis actually lives

Saying “revenue beat by 2,800” is arithmetic. The analysis begins when you ask what’s inside that number, because the headline almost always hides a more interesting story.

Take a single product line:

The headline is +2,800, Favourable. Looks like a clean win. But split it into the two things that moved, and the win develops a crack:

• We sold more units than planned. That lifted revenue.
• But we sold them at a lower price, and that gave some back.

One favourable driver, one unfavourable driver, hiding inside one cheerful-looking total. Suddenly it’s a different management conversation: did we buy that volume with a discount? Is the discount a one-off or is it now structural? Volume “looks great” right up until you decompose it, and discover the margin you traded away to get it.

This price-and-volume split is the simplest decomposition there is. Real ones go further: separating mix (selling more of the low-margin product) from pure volume, splitting cost variances into rate and efficiency, and assembling the whole thing into a bridge that walks from plan to actual one driver at a time. Each of those is a craft worth its own walkthrough, and we’ll give them one. The principle underneath them never changes:

Hold one factor still, flex the other, attribute the gap.

A note on not analysing everything

A discipline hides inside the discipline: don’t investigate every variance. A small wobble on a large line is noise, and chasing it is how an analyst burns a week to explain nothing. Mature teams set a threshold (a size, a percentage, or both) and only dig into what clears the bar. Where to draw that line, and why netting variances against each other quietly hides news, is a topic we’ll take up separately. For now: spend your judgement where the money is.

Where variance analysis runs out

Here’s the honest limit, and it’s the bridge to everything that comes next.

When you split a revenue miss into price, volume and mix, you’ve attributed the gap, but every one of those buckets is itself a number that demands a why. Price fell. Why? Volume rose. Why? Decomposition tells you which lever moved. It does not tell you what moved the lever.

That’s the wall variance analysis hits. It’s a comparison and an attribution. It tells you what differed, with real precision. What it can’t tell you, on its own, is what actually happened in the business to cause the difference: was the volume lift our promotion, or the season we’d have captured anyway? Was the price drop a deliberate decision, or currency we don’t control? Separating what differed from what happened, and reasoning about cause rather than coincidence, is a different mode of analysis, and it’s where this series goes next.

Variance analysis is the indispensable first step. It points the finger at the right driver and turns “why did we miss?” into “why did price fall?”, a question precise enough to actually answer. It just doesn’t answer that last question itself.

In Novi, it’s already there

For most teams, the hardest part of variance analysis isn’t the concept. It’s the labour. Someone rebuilds the comparison every month: pulling actuals, lining them up against budget, flipping signs so favourable reads positive, splitting the drivers, assembling the bridge. It’s a monthly project built on a spreadsheet that breaks the moment a dimension changes.

In Novi, variance analysis is a virtual dimension built into your data tables. You don’t rebuild anything. Actual, Budget, the variance, the percentage variance, the Favourable/Unfavourable view: they’re a dimension you pivot on, the same way you’d swap Products for Regions. Flip from Actual to vs Budget with a drag. The comparison that used to be a monthly chore is simply there, live, on every table that has the data.

Which means the analyst’s time lands where it belongs: not on building the comparison, but on answering the question the comparison raises. Why. And that, as the next article argues, is the half of the job that actually moves the business.