A sales forecast is a prediction of future revenue over a defined period. Done well, it drives hiring, inventory, cash planning and the numbers you give the board. Done badly, it quietly poisons every downstream decision — you staff for demand that never arrives, or run out of stock in your best quarter. The difference between the two is rarely the sophistication of the maths; it is the method’s fit to the question and the discipline behind it.
This guide walks through the main sales forecasting methods, when each one earns its place, and how to combine them into a forecast people actually believe. It is written for finance and commercial leaders who have to commit to a number, not just admire a chart.
Key takeaways
- Forecasting methods split into qualitative (judgement-based) and quantitative (data-based) — the best forecasts blend both.
- Match the method to the situation: a new product needs judgement and analogues; a mature line with years of history suits time-series models.
- Accuracy comes from process — a single owner, a regular cadence, and honest measurement of forecast error — more than from any one technique.
- A forecast without a feedback loop is a guess; track error every period and feed it back into the next cycle.
Qualitative forecasting methods
Qualitative methods rely on informed judgement. They are indispensable when you lack history — new products, new markets, structural change — or when human insight (a known contract, a competitor exit) trumps the data.
Sales-force composite
Aggregate the estimates of the people closest to customers — your reps and account managers. Rich in ground truth, but prone to bias: sandbagging to make targets easy, or optimism to please management. Useful, provided you correct for the systematic bias each person carries.
Expert opinion and the Delphi method
Structured input from people who know the market. The Delphi method formalises this: experts forecast independently and anonymously over several rounds, converging without the loudest voice dominating. Slower, but powerful for uncertain, strategic horizons.
Market research and analogues
For genuinely new offerings, base the forecast on customer research and the launch curves of comparable products. Imperfect, but far better than extrapolating from a history that does not exist.
Quantitative forecasting methods
Quantitative methods use historical data and statistical relationships. They shine where the past is a reasonable guide to the future — stable, high-volume, repeatable demand.
Time-series methods
These project a series forward from its own history, decomposing it into trend, seasonality and noise:
- Moving average — smooths short-term noise; simple and robust for stable demand.
- Exponential smoothing — weights recent periods more heavily; variants (Holt, Holt-Winters) handle trend and seasonality.
- ARIMA — models the autocorrelation in a series; powerful with enough clean history and a careful hand.
Causal and regression methods
Causal models link sales to drivers — price, marketing spend, pipeline, macro indicators, weather. A regression quantifies each relationship so you can forecast sales and answer ‘what if we cut price 5%?’. This is where forecasting meets driver-based planning, and where machine learning increasingly adds value on large, multi-variate datasets.
Pipeline and bottom-up methods
For B2B, build the forecast from the pipeline: opportunities weighted by stage probability and expected close date. Only as good as your CRM hygiene — but close to the commercial reality of the business.
| Method | Best for | Data needed | Watch out for |
|---|---|---|---|
| Sales-force composite | B2B, relationship-led sales | Rep estimates | Optimism / sandbagging bias |
| Time-series (smoothing, ARIMA) | Stable, high-volume demand | Several years of clean history | Structural breaks the model can’t see |
| Causal / regression | Demand driven by known levers | Driver + sales history | Spurious correlation, overfitting |
| Pipeline-weighted | B2B with a defined sales cycle | Clean CRM pipeline | Stale or inflated opportunities |
The right forecasting method is the simplest one that captures how your demand actually behaves. Complexity you can’t explain to the board is a liability, not a feature.
Choosing the right method
There is no single best method — only the best fit. Ask three questions: How much relevant history do I have? How stable is the underlying demand? And what decision does the forecast need to support? A cash forecast for the next quarter, a capacity plan for next year and a five-year strategic outlook call for different tools and different tolerances for error.
In practice the strongest forecasts are blended: a statistical baseline that removes human noise, overlaid with structured judgement for the things the model cannot know. The statistics keep you honest; the judgement keeps you relevant.
A worked example
A subscription business forecasts next quarter’s new revenue. The statistical baseline — Holt-Winters on three years of bookings — projects £1.20m, reflecting trend and Q-end seasonality. The sales team, via a pipeline-weighted view, sees two large deals likely to land, worth £0.25m above the baseline. Finance knows one major customer is churning, removing £0.10m.
The committed forecast becomes £1.20m + £0.25m − £0.10m = £1.35m — a statistical baseline, adjusted transparently for two things the model could not see. Crucially, each adjustment is documented, so next quarter you can check whether the deals landed and the churn bit, and recalibrate.
Measuring forecast accuracy
A forecast you never grade will never improve. Track error every period using a consistent metric — commonly MAPE (mean absolute percentage error) or bias (are you consistently high or low?). Persistent bias is more dangerous than random error, because it means a fixable flaw in your assumptions. Feeding error back into the next cycle is what separates a forecasting process from a forecasting habit, and it is the foundation of a credible rolling forecast.
Turn forecasting from an event into a capability
Metapraxis helps finance teams build rolling, driver-based forecasts that blend statistical rigour with commercial judgement — and that leadership actually trusts.
See our rolling forecasting approach →Sales forecasting techniques vs methods
The terms are often used interchangeably, and in practice the difference is small: a sales forecasting method usually refers to the overall approach (qualitative versus quantitative, top-down versus bottom-up), while a sales forecasting technique refers to the specific mechanic used within it — exponential smoothing, regression, opportunity-stage weighting and so on. What matters is not the label but matching the approach to your data, sales cycle and the decision the forecast informs.
Sales forecasting tools and software
Most teams outgrow spreadsheets once forecasts need to combine CRM pipeline data with finance actuals and run multiple scenarios. The main categories of sales forecasting tools are:
- CRM-native forecasting (e.g. pipeline and opportunity-stage views inside your CRM) — good for sales-led, bottom-up forecasts close to the deal.
- FP&A and EPM platforms — connect CRM pipeline to financial drivers, support driver-based and scenario forecasting, and reconcile the sales forecast to the wider financial plan.
- Statistical and ML tools — apply time-series and causal models to historical demand where data volume justifies them.
The right tool is the one that lets you blend methods, keep one version of the numbers, and update the forecast as a routine rather than a monthly fire drill.
