Variance Analysis

Variance analysis is the comparison of an actual result to a benchmark and the explanation of the gap. The benchmark can be a budget, a forecast, the prior period, the same period last year, or a standard cost. Choosing it is the first decision and it changes what the analysis means: a variance to budget is a question about planning, a variance to prior year is a question about the business, and a variance to standard is a question about operations.

It overlaps with flux analysis without being the same thing. Flux analysis is the close-time review of period-over-period fluctuations in ledger accounts, performed to catch errors before the statements go out. Variance analysis is the broader management and audit technique of explaining differences against any benchmark, including forward-looking ones. A close process usually needs both, and the useful distinction is purpose: flux is a detective control, variance is a management explanation.

Thresholds should be set in two dimensions before any work starts. A percentage threshold alone flags every small account that moved from 100 to 300. An absolute threshold alone ignores a 90% collapse in a small but strategically important account. Requiring both to be exceeded produces a workable population. Setting thresholds after seeing the results, which is common, means the analysis explains what someone already decided was interesting.

Decomposition is what separates analysis from description. A revenue variance splits into price and volume effects, and often into mix, before it is explained. A labor variance splits into rate and efficiency. A cost variance driven by a vendor increase behaves differently from one driven by consumption, and the response to each is different. Netting is the enemy here: a flat total can hide a large favorable variance sitting on top of a large unfavorable one, and the total tells the reader nothing.

The output is the explanation, and most explanations are not usable. "Expenses increased" restates the number. "Professional fees increased $180,000 because the ERP implementation began in March, per the signed statement of work" is specific, quantified, and evidenced. The test is whether a reviewer can verify the explanation without going back to the preparer. A good explanation names the driver, quantifies its contribution, and points at the support. Where several drivers contribute, they should be listed with amounts that sum to the variance rather than gestured at collectively.

Two practices make the whole exercise more valuable over time. Carry explanations forward, so a recurring variance is recognized as recurring rather than re-investigated each period. And close the loop on forecast variances by feeding what actually happened back into the next forecast, which is the only mechanism that makes the benchmark better.

Related terms: Flux Analysis, Month-End Close, General Ledger, Materiality

Related skill: Return year-over-year variance

Go deeper: Accounting Automation

Producing the variance table is fast. Explaining every line with evidence attached is what takes the week. Adopt's agents assemble the drivers from the underlying detail and draft the explanations for review. Sign up free.