3 Demand Forecasting and Planning

Learn how demand patterns and forecasting methods support planning, how accuracy and bias are measured, and how forecasts guide operational decisions under uncertainty.

Demand patterns and forecasts

estimates how much customers are likely to buy, where, and when. It gives supply-chain teams a forward-looking basis for decisions, but it is not a guarantee of future sales. coordinates these estimates with procurement, production, inventory, warehousing, and distribution plans.

Recognizing demand patterns

Past demand can exhibit several patterns:

Level: Demand fluctuates around a relatively steady average.
Trend: Demand rises or falls over time, perhaps as a product gains popularity or declines.
: Demand regularly changes with the calendar, such as higher school-supply sales before a new school year.
Cyclical variation: Demand moves over longer periods, often with economic or industry conditions, but without a fixed seasonal schedule.
Irregular variation: Unusual events, such as a disruption or one-time order, cause changes that are difficult to predict.
: Many periods have little or no demand, interrupted by occasional orders; this is common for spare parts.

These patterns can occur together. For example, a product may have a growing trend and a recurring holiday peak.

Choosing a forecasting method

Choose a forecasting method according to the available data, the forecast horizon, and the decision the forecast must support.

Qualitative judgment uses input from sales teams, customers, or experts. It is useful when historical data is limited, as with a new product, but assumptions and should be reviewed.
Naïve forecasts use a recent actual value as the next forecast and provide a simple benchmark.
Moving averages smooth short-term fluctuations by averaging recent periods. If demand over the last three months was 90, 105, and 111 units, the three-month moving-average forecast is:

\frac{90+105+111}{3}=102\text{ units}

uses past demand while giving more weight to recent observations. Variants can account for trend and .
Causal models, such as regression, estimate demand using factors that may help explain it, for example price, promotions, or weather. A relationship in historical data is not automatically proof of causation.
Combined forecasts bring statistical estimates together with informed adjustments, such as a confirmed promotion or a known customer contract.

A practical approach is to compare candidate methods against a simple benchmark, then use a method that performs reliably for the relevant product, location, and time horizon. Changes in promotions, distribution, or market conditions may make past demand less representative of future demand.

Measuring forecast accuracy

Forecast accuracy compares forecasts with actual demand after the period has passed. Let $A$ denote actual demand, $F$ the forecast, and $n$ the number of periods evaluated.

The , also called mean absolute deviation (MAD), measures typical error in units:

\text{MAE}=\frac{\sum |A-F|}{n}

This unit-based measure can help clarify operational impact. The expresses error as a percentage:

\text{MAPE}=100\times\operatorname{mean}\left(\frac{|A-F|}{A}\right)

MAPE is undefined when actual demand is zero and can be misleading when actual values are very small.

is the average signed error and indicates whether forecasts tend to be systematically high or low. With error defined as actual demand minus forecast, positive means forecasts tend to be too low, or under-forecasted; negative means they tend to be too high, or over-forecasted.

For example, if actual demand is 100 units and the forecast is 90, the absolute error is 10 units and the percentage error is 10%. The signed error is $100-90=10$ , indicating under-forecasting under this convention.

No single metric tells the whole story: a forecast can have small average errors but still be consistently biased. Compare forecasts with actuals over several periods, choose metrics that suit the data, and examine results by product or location as well as in total.

Turning forecasts into plans

Forecasts translate expected demand into operational choices. A higher forecast may lead planners to secure materials earlier, reserve production capacity, adjust workforce plans, or position more inventory near customers. A seasonal increase may also require extra warehouse space or transportation capacity. Lower expected demand can prompt teams to defer purchases or production and avoid excess stock.

Planning decisions should consider forecasts alongside supplier and transportation lead times, available capacity, inventory on hand, and the cost of shortages versus excess inventory. Because forecasts are uncertain, planners can review alternative scenarios, apply appropriate buffers, and update plans when meaningful new information arrives.

Demand and supply planning are linked: forecasts inform supply decisions, while supply constraints may require teams to agree on how to respond to expected demand.