Revenue forecasting is where most business forecasts begin and where most go wrong. The good news: a handful of methods cover almost every situation, each with clear strengths and weaknesses. Knowing which to use when, and where each misleads, is most of the discipline.

Five methods, in order of complexity

1. Trend extrapolation

Take recent growth rate, project forward. "We grew 4% MoM for six months; assume 4% MoM continues." Simple, sometimes useful for stable businesses, dangerously misleading when the underlying dynamics are changing.

2. Top-down

Start with a market or growth-rate assumption, work down. Useful for strategic planning and new-venture scoping. Often too optimistic for operational use.

3. Bottom-up

Start with current activity, build the future from it. For a subscription business: starting MRR + (new MRR from expected new customers) − (churned MRR) − (contraction MRR) + (expansion MRR). For a service business: existing client revenue + new client revenue × probability + retainer renewals. Usually the most accurate method for short and medium horizons.

4. Driver-based

Identify the 2-3 metrics that mathematically produce revenue, forecast each one, multiply. Examples:

  • E-commerce: visitors × conversion rate × average order value
  • SaaS: customers × average revenue per customer
  • Agency: billable consultants × utilization × hourly rate × hours/month
  • Retail: foot traffic × conversion × basket size

Most diagnostic method - when reality differs, you can identify which driver was off.

5. Pipeline-based

For businesses with clear sales pipelines (B2B with multi-touch cycles), forecast revenue by multiplying expected deal value by stage-specific close probability:

  • Lead: 5%
  • Qualified: 15%
  • Proposal: 35%
  • Negotiation: 70%
  • Verbal yes: 90%

Refine the probabilities from your own historical close rates, not the standard numbers. Most accurate method for B2B with sales cycles longer than a month.

When to use each

Method selection by situation
Best when
Trend extrapolationBusiness is stable and dynamics haven't changed
Top-downStrategic planning, new venture, market entry
Bottom-upMost short and medium horizons, especially without a sales pipeline
Driver-basedYou want to diagnose what's working and what isn't
Pipeline-basedB2B businesses with clear multi-stage sales cycles

Combine methods

The best forecasts use multiple methods - one as the primary, others as sanity checks. A pipeline forecast for next quarter checked against a bottom-up build. A driver-based forecast for the year checked against trend extrapolation.

When two methods strongly disagree, that's a signal - something in your assumptions is wrong. Diagnose before forecasting.

Customer-level vs aggregate forecasting

A useful rule of thumb: top 10-20% of customers individually, the long tail in aggregate. Major accounts deserve direct attention - their churn or expansion materially affects the forecast. Smaller accounts are noise individually but matter together.

Common mistakes

1. Optimism on conversion timing

Deals close later than expected. New customer ramp-ups take longer. Build the delays into the forecast.

2. Forecasting in straight lines

Real revenue has seasonality, calendar effects, lumpy deals. A straight-line forecast will always be wrong somewhere.

3. Ignoring churn or contraction

Especially in subscription businesses, growth forecasts often ignore the customers leaving. Net growth is what matters, not gross.

4. Treating pipeline as committed revenue

A pipeline is a forecast input, not a commitment. Stage probabilities exist for a reason. Don't multiply by 1.0.