How to Forecast Sales: Drivers, Methods, and Practical Approaches

Most sales forecasts fail because they answer only one question: "How much will we sell?" A cash flow forecast requires three additional questions: How certain is the sale? When will it happen? When will the cash actually reach the bank? This guide shows how to build a sales forecast that finance teams can use for reliable 13-week cash flow forecasting.

For the first weeks of any forecast, the preparer can review the accounts receivable ageing list and estimate — based on collection probability — how much cash will arrive each week. But for weeks further out, the process is different. First, the forecaster must estimate how much will be sold. Then, based on those sales, the forecaster estimates when and how much cash the company will receive.

This depends heavily on the type of sales a company makes. A business with retail customers collects cash instantly. A business selling on credit terms may wait 60 or 90 days. To forecast receipts correctly, the forecaster must first understand the sales model.

Types of sales models

Before forecasting sales, the forecaster needs to understand what type of sales the business makes. Different sales models generate revenue in fundamentally different ways, and each requires a different forecasting approach.

The sales model determines not only how to forecast volume but also when cash will arrive after the sale. That second question — the timing of cash collection — is covered in detail in the companion post: How to forecast cash receipts from customers.

What creates sales?

Every business has a small number of variables that determine how much it sells. These are called sales drivers. To forecast sales, the forecaster first needs to identify the drivers. Each driver is then forecast separately.

The formula is always the same in structure: volume × price. The volume part changes depending on the industry.

The value of identifying drivers is this: instead of guessing a total sales number, the forecaster forecasts each driver separately. That makes the forecast more transparent and easier to challenge.

Gross vs. net sales: accounting for returns and allowances

A mistake in cash forecasting is taking the gross sales number and assuming it all hits the bank. In reality, customers return goods, demand refunds, and negotiate allowances. A forecaster must adjust gross sales down to net cash receipts.

The cash flow trap of gross sales

If an e-commerce business forecasts €100,000 in sales for the week, but historically 8% of those goods are returned and another 2% are kept by the customer at a discounted allowance due to minor damage, the actual cash retained is only €90,000. If the forecaster uses the gross €100,000 figure, the cash flow forecast will be overstated by €10,000, potentially masking a looming liquidity shortfall.

Types of sales deductions

When bridging the gap between gross sales and net cash, the forecaster must account for four main types of deductions:

• Returns and refunds — Physical goods sent back by the customer. This is highly prevalent in e-commerce (especially apparel) and wholesale distribution.
• Sales allowances — A reduction in the price granted to the buyer after the sale, usually because goods were slightly damaged, delivered late, or did not perfectly match the specification, but the customer agreed to keep them.
• Chargebacks — When a customer disputes a charge with their credit card provider or payment platform. The payment processor forcibly reverses the transaction, often adding a penalty fee on top of the refunded amount.
• Volume rebates and early payment discounts — Agreed deductions. A customer might receive a 2% discount for paying within 10 days, or a 5% end-of-year rebate if they hit a certain purchasing volume.

Sales forecasting approaches

There is no single correct method. The right approach depends on the business, the available data, and how far ahead the forecast needs to reach.

Historical trend forecasting

The forecaster looks at past sales and projects the same pattern forward.

Strength: simple and fast. Works well when the business is stable.

Weakness: assumes the future will look like the past. Breaks down during disruptions, recessions, or rapid growth.

Seasonality forecasting

The forecaster identifies repeating patterns within the year — busy months, quiet months — and applies them to the forecast.

Strength: captures predictable fluctuations. Essential for retail, tourism, and agriculture.

Weakness: requires at least two to three years of history to identify patterns reliably. Unusual years distort the baseline.

Budget-based forecasting

The forecaster uses the company's annual budget as the starting point and spreads it across weeks.

Strength: aligned with management expectations and financial targets.

Weakness: budgets are often optimistic. A budget is a target, not a prediction. Using it as a forecast without adjustment can be misleading.

Driver-based forecasting

The forecaster breaks sales down into its components — volume, price, conversion rate — and forecasts each one separately, then multiplies them together to get total sales.

Strength: transparent and logical. Easy to see which assumption is driving the result.

Weakness: requires good data on each driver. More complex to build and maintain.

Pipeline, order-book, and contract-based forecasting

These three methods all forecast from sales that already exist in some form — a CRM opportunity, a confirmed order, or a signed contract — rather than from projected new business. Pipeline forecasting weights CRM opportunities by their close probability; order-book forecasting uses only confirmed orders; contract-based forecasting uses signed agreements with known delivery schedules and milestones. All three are reliable for what they cover, but none extend past the sales already on the books — new business still has to be forecast separately. The mechanics of each, including how to weight a pipeline and where to find contract-based cash flows, are covered in Forecasting existing sales below.

Statistical forecasting

The forecaster applies mathematical models — moving averages, regression, or machine learning — to identify patterns and project them forward.

Strength: handles large datasets well. Can identify patterns that humans miss.

Weakness: requires clean historical data and technical expertise. Results can be difficult to explain to non-technical stakeholders. Also gives false confidence when data quality is poor.

Forecasting existing sales

This is the most reliable part of any sales forecast. The forecaster is not guessing — the sales already exist in some form. The only task is to find them, organise them, and translate them into a weekly cash forecast.

Order book analysis

The forecaster starts by asking three questions: what orders already exist, what contracts are signed, and what deliveries are scheduled?

These are not forecasts. They are facts. Every confirmed order, signed contract, and scheduled delivery represents a known future sale. Together they form the order book. The forecaster reviews three sources:

• Open sales orders — orders placed by customers that have not yet been delivered or invoiced. These are commitments. The customer has agreed to buy. The business has agreed to deliver.
• Backlog analysis — the total value of work that is contracted or ordered but not yet completed. A large backlog means future revenue is already secured. A shrinking backlog is an early warning sign.
• Contract review — long-term contracts often contain delivery schedules, invoicing milestones, and payment terms. The forecaster reads each contract and extracts the expected cash flows week by week.

The output is a sales forecast built entirely from existing commitments — no assumptions, no estimates, no guesses.

Sales pipeline analysis

Not all future sales are confirmed. Some are still in negotiation. Others are at an early stage. The forecaster needs to assess these opportunities and decide how much weight to give each one.

Most businesses track opportunities in a CRM system. Each opportunity sits at a stage in the sales process. The further along the stage, the more likely the sale will close.

The forecaster multiplies the value of each opportunity by its close probability. This produces a probability-weighted sales forecast.

The forecaster does not use the full €630,000. The probability-weighted total of €469,500 is the more honest estimate.

Other sources of existing sales

The order book and pipeline are the two main sources, but they are not the only ones.

• Recurring contracts — a maintenance agreement or service contract generates predictable monthly revenue. The sale does not need to be "won" each time. It renews automatically or on a fixed schedule.
• Subscription billing — a SaaS or telecom business knows exactly how many active subscribers it has today. Next month's revenue from those subscribers is already largely known, minus expected churn.
• Framework agreements — a supplier may have a framework agreement with a large buyer (common in public sector procurement). Individual orders are placed under the framework. The total is not guaranteed, but historical order patterns give a strong indication.
• Retention releases — in construction and engineering, retention money is held back until project completion. The release date is often contractually defined. This is a known future cash receipt.
• Agreed but unsigned contracts — sometimes a deal is fully agreed and only awaiting legal sign-off. Technically not in the order book yet, but the forecaster may treat it as near-certain.

Forecasting new sales

Once the forecaster has extracted everything from the order book, pipeline, and recurring contracts, there will still be gaps in the 13-week forecast. These gaps need to be filled with estimates of new sales — sales that have not yet been won or ordered. This is where uncertainty begins.

Historical analysis

The starting point for forecasting new sales is historical data. The forecaster asks: what did this business actually sell in previous periods, and does it follow any repeating pattern?

Weekly and monthly sales history shows the baseline. The forecaster looks at the last two to three years of sales data and identifies the average level of activity.

Seasonality is the most important pattern to identify. Most businesses are not flat throughout the year. A garden equipment retailer sells more in spring. A logistics company moves more volume before Christmas. A construction firm slows down in winter.

To build a seasonal sales profile, the forecaster calculates what percentage of annual sales fell in each month historically. This percentage is then applied to the current year's expected annual sales.

If annual sales are expected to be €10 million, the forecaster allocates €600,000 to January, €900,000 to April, and so on. This becomes the seasonal baseline for new sales.

Two important caveats. First, unusual years — a pandemic, a major contract win, a market disruption — distort the baseline. The forecaster should exclude or adjust for these. Second, historical patterns only hold if the business model has not changed significantly.

Management forecast review

Historical analysis gives a baseline. But management often has information that history cannot capture — a new product launch, a large tender in progress, a market they are entering or exiting.

The forecaster collects and documents forecasts from the people closest to the business:

• Country manager forecasts — in a multinational, each country manager typically prepares a local sales forecast. They know their customers, their pipeline, and their market conditions better than anyone at group level.
• Sales director forecasts — the sales director tracks the pipeline and has a view on which deals will close within the 13-week window.
• Business unit forecasts — in a diversified business, each unit may have its own forecast based on its own drivers and customer base.

The forecaster does not simply accept these forecasts. The first task is to understand how each forecast was prepared — whether it is based on the order book, the pipeline, last year's numbers plus a growth assumption, a budget target, or gut feeling. A forecast based on confirmed orders is very different from a forecast based on a budget target set eight months ago.

Forecast confidence decreases over time

Confidence naturally declines as the forecast horizon extends. What starts as near-certain facts in the immediate weeks gradually shifts toward educated assumptions further out.

Don’t treat the entire 13-week period with uniform certainty. Adjust expectations and risk buffers accordingly.

This awareness prevents overconfidence in distant weeks and helps decision-makers understand where the real risks in the cash flow forecast lie. When building your own 13-week model, always ask: “How many of these numbers are facts versus assumptions?”

Forecast quality assessment

Collecting management forecasts is only half the work. The other half is deciding how much to trust them.

Evaluating forecast providers

The forecaster assesses the reliability of each forecast provider (country managers, sales directors, business units, etc.). This is done by comparing past forecasts with actual results once sufficient history is available.

Note for new forecasters or new providers: When there is no historical track record (e.g., a new forecaster starting or a new country manager), it is impossible to calculate accuracy metrics initially. In such cases:

• Use conservative default assumptions or industry benchmarks.
• Apply higher scrutiny and require detailed supporting evidence for the first few periods.
• Start tracking forecasts from the very first submission so that a reliable track record can be built quickly.

Once at least 3–4 periods of history exist, three key metrics are calculated:

• Forecast accuracy % — measures how close the forecast was to actual sales on average.
  Formula: (1 − |Actual − Forecast| / Actual) × 100
  * The vertical bars | | mean absolute value (we ignore whether the forecast was higher or lower than actual and only care about the size of the difference).
  Example: A forecast of €1,000,000 against actual sales of €900,000 gives 88.9% accuracy.
• Mean error — shows the average size of the error, regardless of direction. A mean error of €50,000 means the forecast was typically off by €50,000 in either direction.
• Bias — the most revealing metric. It reveals whether a forecaster systematically over- or underestimates.
  Formula: Average (Forecast − Actual)
  Positive bias = consistent overestimation. Negative bias = consistent underestimation.

Track forecast vs actual per country/manager over time. For example:

The forecaster uses this scorecard to apply adjustments to incoming forecasts before incorporating them into the 13-week cash flow. For example, if the France country manager has historically overstated sales by ~20%, the forecaster applies a corresponding downward adjustment.

Advanced tips for better sales forecasts

Even the best forecasting methods can be undermined by overly optimistic assumptions. These practical techniques help validate and strengthen the sales forecast.

Historical run-rate comparison

Pull the last four to six quarters of actual sales by country, business unit, or entity. Calculate the average weekly or monthly run rate and the typical range. If management is suddenly forecasting 15% growth, ask for specific evidence — a new contract, product launch, or confirmed order. If they cannot point to concrete drivers, treat the uplift with scepticism and consider using a more conservative baseline.

Budget comparison

Compare the submitted forecast against the approved annual budget. Ask: what has changed since the budget was set? Look for credible explanations such as new customers or a confirmed tender win. If the forecast is simply "budget plus optimism," it needs adjustment. Budgets are targets, not predictions.

Capacity and operational cross-check

Sales forecasts must be physically deliverable. If a business unit is forecasting 30% growth, verify whether it has the necessary staff, production capacity, inventory, or supply chain to support it. A forecast that exceeds operational reality is not a forecast — it is wishful thinking.

Market reality check

Step outside the internal view and assess external conditions. Are major customers facing bankruptcy? Is there an economic slowdown, a new competitor, or a supply chain disruption? If the broader market is contracting but the forecast shows strong growth, demand compelling evidence.

Build a parallel independent estimate

Take management's forecast and build a separate version using historical run rates plus only the verifiable uplifts that can be confirmed — signed contracts, confirmed orders. Present both side by side. The gap between the two numbers clearly highlights the uncertainty range.

Publish the scorecard

The accuracy and bias scorecard described earlier under forecast quality assessment works best when it is visible, not kept as an internal note. Circulating the figures back to country managers, sales directors, and business unit heads — even informally — creates pressure to forecast more carefully, since everyone knows their track record is being measured. Businesses that publish these scorecards consistently see forecast bias narrow over two or three cycles, simply because providers start defending their numbers before submitting them.

Accurate sales forecasts are only half the battle. Sales forecasts must be converted into reliable weekly cash flow data (read How to Forecast Cash Receipts from Sales).