Asking your sales team for a revenue forecast is like asking the weather to predict itself. You need a model based on historical data, pipeline metrics, and external factors, not optimistic guessing. A good sales forecast does not need to be perfect. It needs to be better than gut feel, and even a basic data-driven model will outperform intuition by a wide margin.
Gather Your Historical Data
Export at least 12 months of sales data from your CRM or accounting system. You need date of sale, amount, and ideally the source of the deal and how long it took to close. If you have pipeline data, deals in progress with stages and expected values, that adds another dimension. Do not worry about messy data; start with what you have. Twelve months of imperfect data is more useful than no model at all.
Start with a Simple Time-Series Model
The simplest useful forecast is a time-series model that identifies trends and seasonality in your historical revenue. Facebook's Prophet library (now Meta's, open source, free) makes this surprisingly easy. Feed it your monthly revenue data and it produces a forecast with confidence intervals. It automatically detects yearly and weekly patterns, holiday effects, and overall growth trends. For many small businesses, this is enough.
Add Pipeline Weighting for Accuracy
If you have a sales pipeline, weight each deal by its probability of closing based on the stage it is in. Proposals submitted might close at 40%, verbal agreements at 75%. Multiply each deal's value by its probability and sum them for a pipeline-weighted forecast. Compare this against your time-series forecast and use the average. Two models combined almost always outperform either one alone because they capture different signals.
Incorporate External Factors
Some businesses are affected by factors outside their control, seasonality, weather, economic indicators, competitor activity. Add these as features to your model if you can quantify them. A landscaping company's revenue correlates with temperature and precipitation. An insurance agency's new business correlates with home sales in their area. Even adding one or two relevant external factors can improve forecast accuracy by 15-25%.
Tracking Forecast Accuracy Over Time
Measure your forecast accuracy by comparing predictions to actuals each month. Calculate Mean Absolute Percentage Error: the average of how far off each prediction was, expressed as a percentage. A 10-15% MAPE is good for a small business forecast. Below 10% is excellent. If accuracy is poor, look at which months have the biggest errors and investigate what happened. The model improves as you add data and adjust for factors you missed.
Want a custom sales forecasting model for your business? We build models that pull from your CRM and accounting data to predict revenue accurately. Optimization & Analytics
Related industries: Real Estate & Property Sales, Insurance Agencies & Brokerages, Staffing Agencies & Recruiting Firms, E-commerce & Online Retail, Home Services (HVAC, Plumbing, Electrical)