Every month, finance teams around the world produce a table showing budget versus actuals. And every month, the same thing happens: the numbers are reviewed, the variances are noted, and the meeting moves on. Nobody changes anything. Nobody learns anything. The budget misses again the following month, for different reasons.
Variance analysis done badly is compliance theatre. Variance analysis done well is one of the most powerful feedback loops in a business — a systematic way of learning what your financial model gets wrong, correcting it, and improving the quality of decisions over time.
This guide explains how to do it properly.
What variance analysis is and what it's actually for
Variance analysis is the process of comparing actual financial results to a plan (usually the budget), identifying and quantifying the differences, and understanding the causes. It is one part of the wider planning cycle described in our guide to FP&A for startups.
The word "understanding" is where most companies stop short. They identify the variance — "revenue was €45K below budget" — and note it in the management pack. But the purpose of the analysis is not identification; it is causation and response. Why was revenue €45K below budget? What does that tell us about the reliability of our revenue model? What is the right response?
Variance analysis serves three distinct purposes:
1. Accountability and performance management. The primary function: holding the business (and individuals within it) accountable for commitments made in the budgeting process.
2. Model calibration. Every variance contains information about what the financial model assumed incorrectly. Systematic variances — the same type of miss, quarter after quarter — signal a structural flaw in the model's assumptions that should be corrected.
3. Early warning. Some variances are the first visible signal of a trend that will become significant. Gross margin slipping 1.5 percentage points below budget in month 3 is recoverable. The same slip unaddressed for six months creates a structural problem. Catching it early is the value.
The four categories of variance: a taxonomy that changes what you do with the information
Not all variances are the same, and treating them the same leads to poor responses. A useful classification:
Timing variances: The revenue or cost will arrive — it's just in a different period than planned. An enterprise deal expected to close in December signs in January. Hosting costs billed quarterly hit in a different month than the monthly model assumed. Timing variances are real but not indicative of business performance problems. The appropriate response is to update the forecast to reflect the timing, not to treat it as a miss.
Volume variances: More or fewer units sold, more or fewer customers acquired, more or fewer headcount hired than planned. Volume variances tell you whether the operational execution of the plan is on track — if you planned to acquire 15 new customers and acquired 11, that's a volume variance in new customer acquisition.
Price/rate variances: Same volume, different price. Average contract value lower than budgeted because of discounting. Cloud infrastructure costs higher than budgeted because of on-demand pricing rather than committed use discounts. Price variances often signal either a commercial execution issue (discounting culture) or a cost management issue (overpaying for suppliers).
Mix variances: The total revenue number might hit, but the mix of products, geographies, or customer segments contributing to it differs from plan. Mix variances are important because different parts of the revenue mix have different margin profiles — so a mix shift can produce a P&L outcome that looks neutral on topline but is materially different on gross margin and unit economics.
Classifying each significant variance into one of these four categories before writing the commentary changes the quality of the analysis dramatically.
How to write variance commentary that drives action
Variance commentary is the narrative explanation of why actuals differed from the plan. Most variance commentary is written to satisfy a reporting requirement, not to inform a decision. The difference in quality is immediately apparent.
Poor variance commentary: "Revenue was €42K below budget due to delayed enterprise deals."
Good variance commentary: "Revenue was €42K below budget, driven by two enterprise deals (combined TCV €180K) that slipped from December into January due to extended procurement approval timelines following customer year-end finance freezes. Both deals are now signed (January 8th and January 14th) and will recognise in January. Excluding the timing impact, underlying revenue performance was in line with budget across all segments. No change to the Q1 revenue forecast as a result."
The difference: the good commentary identifies the specific cause, quantifies the items driving the variance, explains whether it is recoverable, and states the forward implication. A board member reading this commentary has everything they need to form a view on whether this is a problem or a timing artefact.
Three questions every variance commentary should answer:
- What specifically drove the variance? Name the deals, the cost items, the operational decisions — not just the category.
- Is this variance timing, volume, price, or mix? Each has a different forward implication.
- What is the forward impact? Does this variance change the forecast? Does it require a management response?
Building a variance analysis framework for your management pack
The monthly management pack should include a variance analysis section with the following structure:
The summary table A single-page table showing actuals vs. budget for the month and year-to-date, for each major P&L line. Variances expressed as both absolute (€) and percentage. Items above the significance threshold (typically 10% or €5K, whichever is lower for your scale) flagged for commentary.
The commentary section For each flagged variance: one paragraph maximum per item, structured around the three questions above. Total commentary should take a reader no more than 5–8 minutes to consume.
The forecast implication A separate section that captures: which variances change the full-year forecast, by how much, and in which direction. This section directly connects the backward-looking variance analysis to the forward-looking forecast update — closing the loop between what happened and what comes next.
The meta-analysis: tracking forecast accuracy over time
The most valuable long-run use of variance data is tracking the accuracy of your own forecasts over time — not to be self-critical, but to systematically identify where the model consistently makes the same mistake.
A simple tracker: for each month, record the forecast made 30 days prior and the actual outcome for the 10 most important P&L lines. Calculate the average absolute percentage error for each line over rolling 6-month and 12-month periods.
What this tells you:
- Revenue lines with high forecast error often signal that the leading indicators driving the revenue model (pipeline, conversion rate, deal velocity) are themselves inaccurate or poorly tracked.
- Cost lines with high forecast error often signal that the budget assumptions were based on incomplete information — a supplier contract not yet signed, a hiring plan that assumed different ramp times than reality.
- Lines with consistently negative forecast error (actuals always below forecast) signal systematic optimism bias that should be corrected in the modelling assumptions.
A fractional CFO who tracks this over 12 months will dramatically improve the forecast accuracy of the financial model — because every miss is a learning opportunity, and every learning opportunity is acted on.
FAQs
How significant does a variance have to be before it needs commentary? A common threshold: anything greater than 10% of the budgeted line item OR greater than a fixed absolute amount (often €5,000–€10,000 depending on the scale of the business), whichever is the lower bar. Below these thresholds, the variance is noise. Above them, it warrants explanation.
Should I do variance analysis against budget or against the prior forecast? Both, and for different reasons. Budget vs. actuals measures performance against commitments — it's the accountability mechanism. Forecast vs. actuals measures the quality of your forecasting — it's the model calibration mechanism. Neither tells the full story without the other.
What should I do when variances consistently go in the same direction? If actuals are consistently below budget on revenue (or consistently above budget on costs) for 3+ months, the budget assumptions are wrong and need to be addressed. This is not a question of updating the forecast — it's a question of whether the underlying budget model is calibrated correctly. A persistent negative pattern typically requires a formal reforecast or at minimum a documented assumption revision.
BB Financial Services Kft prepares management reporting packs and variance analysis frameworks for startups and SMBs as part of the fractional CFO engagement. Get in touch if your monthly numbers inform reporting but don't inform decisions.