Driver based
DCFForecasting each line from the operational quantity that actually causes it, rather than as a percentage of revenue.
A driver based model separates volume from price, and links each cost to whatever genuinely moves it. Headcount drives staff costs, square footage drives rent, units drive raw materials.
This is what makes a model useful rather than decorative. Forecasting every cost as a flat percentage of revenue assumes perfect operating leverage neutrality, which is exactly the assumption most worth testing, and it makes the model incapable of showing margin expansion or contraction for any structural reason.
It also makes the model answerable to reality. If gross margin is forecast to rise, a driver based build says whether that comes from price, mix, input cost or scale, and each of those is a separately checkable claim.
The cost is complexity, and the discipline is to build drivers only where they change the answer. A line that is genuinely proportional to revenue should be forecast that way.
Worked example
Forecasting staff costs at 22% of revenue assumes headcount scales perfectly with sales.
A driver based build forecasts 400 employees at an average 65 each, so 26.0 of cost, and lets headcount grow more slowly than revenue.
At 10% revenue growth with headcount up 4%, the percentage of revenue method shows a flat margin while the driver based one shows margin expanding. That difference is operating leverage, and it is the thing most worth modelling.