Driver-Based Planning

Driver-based planning builds a financial plan around the operational drivers (headcount, price, volume, conversion rate) that actually determine outcomes, instead of applying a flat growth percentage to last year’s numbers. The plan becomes a set of stated assumptions plus the arithmetic that connects them, rather than a grid of typed values.

How a driver-based model is built

Start from the output line and work backward until you reach something the business actually manages. Revenue becomes customers times average contract value. Customers become opening customers plus new minus churned. New customers become leads times conversion rate. Support cost becomes tickets per customer times cost per ticket. Each step stops when it hits a number an operator can change or defend.

The result is a driver tree: a small set of inputs at the edges feeding a calculated structure. A useful test is to count the cells someone would need to edit to model 20 percent faster growth. In a driver-based model it is one or two. In a typed model it is most of the sheet.

Where driver-based planning goes wrong

The common failure is over-decomposition. A model with 400 drivers is not more accurate than one with 30; it just has 400 places to be wrong, and nobody maintains the ones that stopped mattering. Pick drivers that are material, that someone owns, and that can be measured against actuals.

The second failure is fake drivers. If sales headcount feeds revenue but the ratio between them was reverse-engineered to produce the number leadership wanted, the model looks driver-based and behaves like a hardcoded plan. A real driver has an observable history you can check against actuals.

The third is stale drivers. Conversion rates and cost ratios drift. Reviewing driver values against actuals each cycle is what keeps a rolling forecast honest rather than merely current.

What good looks like

A good driver model is small enough to explain in a meeting, has an owner per driver outside finance, and reacts sensibly to a shock: cut conversion by half and the downstream lines should move in a way that matches intuition.

Change one driver in plain language and every dependent line in a Rexfin model recomputes deterministically, because the plan is built on drivers rather than static formulas. Each historical value behind a driver is cited back to the source record it came from.