Driver-Based Forecasting of Financials
Instead of growing every line item on a financial statement by a flat percentage, a driver-based model ties each line to the real operational metric that actually causes it to move.
Prerequisites: Revenue Builds: Top-Down TAM vs Bottom-Up Units
The laziest way to forecast a financial statement is to grow every line by the same percentage: "revenue grows 8%, so let's grow cost of goods sold 8%, and rent 8%, and headcount cost 8%." It produces a tidy-looking model, and it is almost always wrong, because none of those things actually move together in reality. Rent doesn't grow with revenue at all until the company needs a bigger building. Headcount cost moves in lumps when new people are hired, not smoothly. Cost of goods sold moves with units sold and input prices, which can diverge sharply from revenue if pricing changes.
Driver-based forecasting replaces the flat-percentage habit with a rule: every line on the model should be projected from the actual operational variable that causes it to change, not from an assumed link to revenue.
A driver is the real-world quantity that causes a financial line to move — units sold, headcount, square footage, machine hours — not revenue itself. Modeling the driver, then letting revenue and costs fall out of it, produces a forecast you can actually challenge line by line.
How a driver-based model is built
Pick the driver for each major line, then build the line from it: revenue from units sold times average price, not from a top-line growth rate; cost of goods sold from units sold times unit input cost, not from a gross margin percentage assumed constant forever; SG&A from planned headcount times average salary, not from a percent of revenue; capital expenditure from planned capacity additions, not from a percent of prior-year sales.
Worked example
A manufacturer forecasts next year using drivers instead of a blanket growth rate. Units sold are projected to grow from 1 million to 1.1 million units based on the sales team's order pipeline. Average selling price is held flat at $50, so revenue = 1.1 million × $50 = $55 million, versus $50 million this year — 10% growth.
Cost of goods sold is driven by a separate assumption: unit cost is expected to rise from $30 to $32 per unit because a key input commodity is getting more expensive. COGS = 1.1 million × $32 = $35.2 million. Gross margin comes out to ($55m − $35.2m) / $55m = 36%, down from 40% this year — a margin compression a flat-percentage model (which would have just held margin at 40% by assumption) would have completely missed, because the real driver of the cost line was commodity price, not revenue.
What this means in practice
Driver-based models take longer to build because every line needs its own defensible assumption, but they are far easier to stress-test: a skeptical reader can challenge "why 1.1 million units?" specifically, instead of being stuck arguing with an opaque blended growth rate. They also make scenario analysis meaningful — flexing the unit-cost driver alone shows the margin sensitivity that a single "grow everything 8%" model can't isolate.
A model that looks driver-based on the surface but quietly still ties SG&A or COGS to "percent of revenue" hasn't actually escaped the flat-growth trap — check that every major line has its own independent driver, not just the top revenue line.
Further reading
- Damodaran, Investment Valuation (ch. 12, forecasting fundamentals)