Variance Analysis

Variance analysis explains the difference between an actual result and a budgeted or forecast figure (price variance, volume variance, timing): the core narrative work of an FP&A team every close. The output is not the number. The output is the explanation of the number, in terms someone outside finance can act on.

How variance analysis is calculated

Start with the total variance: actual minus plan, expressed in currency and as a percentage. Then decompose it. Revenue variance usually splits into price and volume: hold volume at plan and value the price difference, then hold price at plan and value the volume difference. Cost variance splits into rate and usage on the same logic. Whatever is left after the named components is mix, timing, or an error in the decomposition.

Sign convention matters more than it looks. Agree once whether a favorable variance is positive or negative, and apply it the same way to revenue and expense lines, otherwise the summary row stops meaning anything.

Why variance analysis breaks in practice

Four things go wrong repeatedly. Timing gets read as performance: an invoice that slipped a week shows up as a miss, not a shift. The comparison base drifts, because the plan was re-versioned mid-year and nobody recorded which version the report compares against. Materiality is ignored, so the commentary lists twenty small variances and buries the two that matter. And the narrative is retyped in a slide deck, where it stops matching the model the moment the model updates.

A materiality threshold fixes the third problem directly: set a floor in both currency and percentage terms, and only explain what clears it. The version problem is fixed by treating each plan version as an immutable comparison base rather than a workbook that keeps changing.

What good looks like

Strong variance commentary names a cause, quantifies it, and says what happens next: “Services revenue is 180k under plan, roughly two thirds timing on the delayed enterprise implementation, expected to land next quarter; the remainder is lower realized rates.” Every figure in that sentence should be traceable to a source record, and the whole thing should reconcile to the total variance.

Rexfin generates variance narrative from the same cited figures already in the model rather than from manually retyped commentary, so the explanation and the numbers cannot drift apart. Calculations run deterministically against one governed set of definitions, and each contributing actual carries its own audit trail.