Variance analysis that actually tells you something

Plan versus actual is the oldest report in finance, and often the least useful. A grid of green and red numbers tells you that something missed, but not why, and certainly not what to do next. Variance analysis only earns its keep when it explains itself.

The problem with a variance grid

A standalone variance report is cut off from the model that produced the plan. So when revenue lands under budget, someone still has to go and find out whether it was fewer customers, a lower price, or a deal that slipped. The report raises the question and a person answers it, slowly.

Variance where the drivers live

When variance is worked out inside a model built on drivers, the explanation is already there. The model knows the plan assumed a certain number of customers at a certain price, so it can attribute the miss to the driver that actually moved. You get volume was down while price held, instead of revenue was down.

• Attribution: which driver caused the gap, and by how much.
• Accuracy: how good the forecast has been over time, so you trust it more.
• Action: whether this is a one off or a trend that changes the outlook.

The best variance report answers the follow up question before anyone has to ask it.

Built in variance analysis turns the monthly review from a hunt into a conversation. The numbers reconcile, the drivers are named, and the meeting can spend its time on the decision rather than the detective work.