Variance Analysis: How Controllers Turn Numbers into Management Insight

What variance analysis actually is

Variance analysis compares what happened to what you expected - typically actual to budget, actual to forecast, or actual to prior period. A good controller does not just produce the variance numbers. They explain why revenue came in $80K below plan, why marketing spend is up 30% from last quarter, and whether the gross margin drop is a one-time event or a trend. The commentary is the deliverable.

Companies that do variance analysis well treat it as a product. They have a template. They know which variances cross the materiality threshold. They have a standing list of questions for each line item that always needs commentary. The output is consistent month over month so that readers know where to look.

A quick materiality rule that works for most companies: explain any variance greater than the smaller of $10K or 10% of the line item. Below that, commentary adds noise without adding information. Above that, the commentary is expected. Having a written materiality threshold protects the controller from pressure to explain every small movement, which is how variance analysis turns into busywork.

Some variances are structural and should be predicted in the plan, not explained after the fact. If your business has a December revenue spike every year, or a Q1 expense ramp as you hire ahead of the selling season, those belong in the budget. Explaining predictable seasonality as a variance wastes everyone's time. Build the pattern into the plan, then variance analysis only flags the actual surprises.

Revenue variance

Start with volume versus price. Did we sell fewer units, or were the prices lower? Break it down by product line, customer segment, or channel depending on your business. Identify whether the miss was concentrated in one area or broad-based. Note any one-time events - contract terminations, delayed renewals, ramp-up delays. This is what a CEO needs to act on, not just the number.

A useful format: state the variance in dollars, then decompose into drivers. For example, "Revenue was $80K below plan for the month, driven by $50K of delayed renewals (two accounts) and $30K of lower new logo acquisition (Q1 ramp behind plan)." The decomposition is the analysis. Without it, a reader is left guessing what a revenue miss means.

Customer concentration makes revenue variance tricky. If one customer represents 25% of revenue and pushes their renewal from March to April, that is a timing issue, not a business issue. Good variance commentary separates timing from run-rate, one-time from recurring. Investors and board members ask specifically about this distinction, and the controller should be giving them the answer before the question.

Pricing variance matters in B2B businesses but gets overlooked. If you signed 10 deals this month at an average ACV 8% below plan, that is a pricing issue that may or may not reverse. The commentary should call out whether discount rates are trending up, whether it is concentrated in one sales rep or segment, and what the forward view is. Pricing variance today often predicts margin pressure tomorrow.

Expense variance

Payroll variances usually tell a hiring or timing story - a role filled later than planned, a severance payment, a bonus accrual. SaaS and software variances often expose tools nobody is using. Marketing variances need to be linked to campaigns to be meaningful. Every line item with a meaningful variance gets a sentence of explanation.

Expense variances follow a pattern: most months are within 5% of plan on every line. The exceptions are worth writing about. An unexpected $25K in legal fees because of a contract dispute. A $12K bump in cloud costs because of a data ingestion job that ran longer than expected. A missed hire that left $18K of budgeted comp unspent. These are the interesting variances, and the explanations are what the CEO uses to make next-month decisions.

Payroll variance is worth splitting by function because "payroll up $35K" tells you nothing. Is engineering hiring ahead of plan? Is customer success behind? Is G&A bloated? The controller should have a headcount-by-function table that ties to the payroll variance so the commentary writes itself.

Watch for variances that cancel each other out in totals but matter at the line level. You might be on plan overall on OpEx because legal was $20K over and recruiting was $20K under. The netted view makes things look fine. The line-level view shows that legal is consistently over-running and recruiting is consistently under-filled. Both are worth addressing. Neither gets addressed if only the total variance is reviewed.

Turning variance into decisions

The point of variance analysis is not to explain the past. It is to change what you do next. Every variance commentary should end with an implication: we should reforecast, we should hold the spend, we should accelerate hiring, we should investigate further. Without the implication, the analysis is just accounting. With it, the analysis is management.