Driver-Based Forecasting

The lazy approach grows each line by a flat percentage — revenue +8%, COGS +6% — which is fast but a black box: you can't say why growth is 8% or test what happens if pricing slips. Driver-based forecasting decomposes the number into operational levers you can defend and flex. For a SaaS business: revenue = beginning customers × (1 − churn) + new customers, all × ARPU. For a retailer: revenue = number of stores × average sales per store, with a separate same-store-sales growth and new-store build. For a manufacturer: units × price. Each driver can be sourced from management guidance, industry data, or historical trends, making the forecast auditable and connecting it to real-world events (a price increase, a new store opening, a churn improvement).

Driver-based models let you do meaningful sensitivity and scenario analysis — you can ask 'what's the IRR if churn rises 200bps?' which a flat-growth model simply can't answer. They also surface internal contradictions: if you assume 40% volume growth, the model shows whether capacity (capex) and working capital can support it. In diligence, driver-based forecasts let the deal team challenge each assumption independently rather than debating a single opaque growth figure. The downside is they take longer to build and require more data, so analysts reserve full driver builds for the key revenue lines and use simpler ratios for immaterial ones.

Revenue: price × volume, customers × ARPU, installed base × utilization, backlog conversion. COGS and gross margin: cost per unit, input commodity prices, or a margin percentage that itself trends with scale. Operating expenses: often a percent of revenue, or headcount × fully-loaded cost per employee for people-heavy businesses. Capex: percent of revenue, or capacity additions × cost per unit of capacity. Working capital: days sales outstanding, days inventory, days payable outstanding applied to revenue or COGS. The art is choosing the right granularity — enough to be defensible without drowning in detail.