What is forecasting?
Forecasting is the disciplined practice of projecting future financial and operational outcomes — revenue, expenses, headcount, cash flow, and operating KPIs — based on a combination of historical data, leading indicators and explicit assumptions. Forecasting is the bridge between strategy and execution: a clear forecast turns plans into measurable, accountable commitments.
Forecasting horizons
- Weekly cash forecast (13 weeks): rolling weekly view of receipts and disbursements — critical at low runway.
- Monthly operating forecast (quarter and full year): revenue, gross profit, operating expenses, hiring plan, EBITDA.
- Annual plan: formal yearly P&L, balance sheet and cash budget tied to strategy.
- Multi-year (3–5 year) model: for fundraising, M&A or capital allocation — sensitive to long-horizon assumptions.
Common forecasting methods
- Top-down: start with market size and target share; useful for strategy framing, weak for execution.
- Bottom-up: roll up from per-rep quota, per-customer ARR, per-cohort retention — much more accountable.
- Cohort-based: project recurring revenue from each acquisition cohort’s retention curve.
- Regression / driver-based: connect outcomes to specific drivers (marketing spend → leads → conversion → revenue).
- Scenario / Monte Carlo: distribution-based forecasting that quantifies uncertainty.
Forecast accuracy and discipline
The value of a forecast lies less in being right and more in being explicitly wrong — a documented set of assumptions that can be revised as data arrives. Tracking forecast vs. actuals monthly, and recalibrating the drivers, is what makes the forecast progressively better.
- Variance commentary: for every line, explain why actuals differed from forecast.
- Rolling re-forecast: update the forward view monthly as assumptions change.
- Sensitivity table: show how the outcome shifts with key assumption changes (CAC, churn, ASP).
Do: build the forecast bottom-up from observable drivers; document every assumption; track variance monthly and recalibrate.
Don’t: set a forecast and then defend it past the data — treating the forecast as a commitment rather than a tool corrupts learning.