Operations Management and Strategic Management · Operations Planning

Forecasting Demand for Operations: Qualitative and Quantitative Methods

Updated 10 October 2026 · Fact-checked

Demand forecasting is estimating future demand for a product or service so that operations can plan capacity, materials and manpower. Qualitative methods use expert opinion and surveys. Quantitative methods use past data, such as moving averages, exponential smoothing and regression. Choose the method by data availability, time horizon and cost, then check the error.

Understand Forecasting Demand for Operations

Operations cannot be planned without knowing what customers will want. Capacity, inventory, purchasing, staffing and scheduling all depend on expected demand. Demand forecasting is the process of estimating that future demand. A forecast is never exact, but a good one reduces idle capacity, stock-outs and excess inventory.

Forecasting methods fall into two groups. Qualitative methods rely on judgement and opinion. They are used when past data is missing or unreliable, for example for a new product. Quantitative methods use numbers and are used when past data is available and the pattern is likely to continue.

Common qualitative methods are: Delphi method (a panel of experts answers questionnaires anonymously in several rounds, with summarised feedback after each round, until opinions converge); market survey (customers or prospective customers are asked about buying intentions, usually by questionnaire or interview); sales force opinion (salespeople estimate sales in their areas); executive opinion (senior managers pool their views). The key difference: Delphi collects expert opinion in anonymous, repeated rounds, while a market survey collects the views of customers or the market.

Quantitative methods are of two types. Time series methods project past demand into the future. They include the naive method, simple moving average, weighted moving average and exponential smoothing. Causal methods relate demand to other factors, such as price, income or advertising. Regression analysis is the standard example.

No forecast is perfect, so you measure forecast error = actual demand − forecast. Smaller average error means a better method. Short-term forecasts suit scheduling and inventory. Long-term forecasts suit capacity and location decisions.

Key rules to remember

Simple moving average (n periods): Forecast for next period = (sum of demand of last n periods) ÷ n; Every period in the window has equal weight. A larger n gives a smoother but slower-reacting forecast.
Weighted moving average: Forecast = Σ (weight × demand of that period); Weights must add up to 1 (or divide by the sum of weights). Give the highest weight to the latest period unless told otherwise.
Exponential smoothing: Ft+1 = α × At + (1 − α) × Ft; α is the smoothing constant, between 0 and 1. At is actual demand and Ft is the earlier forecast for the same period. A higher α gives more weight to recent demand.
Forecast error: Error = Actual demand − Forecast; A positive error means demand was under-forecast.
Mean absolute deviation (MAD): MAD = Σ |Actual − Forecast| ÷ number of periods; Ignore signs when adding. The method with the lower MAD is more accurate.
Linear regression trend line: Y = a + bX; Y is demand and X is the period or causal variable. a and b are found by the least squares method.

How to solve Forecasting Demand for Operations questions

Use this approach for both theory and numerical questions.

1Read what is asked: a method description, a comparison, or a calculation of a forecast or error.
2For theory, state the meaning first, then the working of the method, then its use and one limitation.
3For a comparison, draw two columns mentally: basis, method A, method B. Cover who gives the input, the data used, the cost and the typical use.
4For a numerical question, note the method named, the window n or α, and the starting forecast if given.
5Write the formula, then substitute the figures in a neat table, period by period.
6If error is asked, compute actual minus forecast for each period, take absolute values, and divide by the number of periods for MAD.
7Compare methods using the error measure and state which is better and why.
8Write the final forecast in a clear line with units.

Quickest way: Fast routine for forecasting sums

When to use it: Use for moving average and exponential smoothing problems in the 14-mark questions and for MCQs.

For a moving average, add only the last n figures and divide by n. Do not use older data.
For exponential smoothing, rewrite as Ft+1 = Ft + α × (At − Ft). It needs one small subtraction.
Check that α was applied to actual demand and (1 − α) to the old forecast.
Sanity check: the forecast should lie between the old forecast and the actual demand.
For MCQs, eliminate options where the weights do not add to 1 or the wrong months were used.

Common mistakes in Forecasting Demand for Operations

Treating Delphi and market survey as the same thing.
Both involve asking people questions.
Fix: Delphi uses a panel of experts, anonymous, in repeated rounds with feedback. A market survey asks customers or the market about buying intentions, usually once.
Using the wrong periods in a moving average.
Students include the current unknown period or more than n periods.
Fix: The forecast for a period uses only the n periods immediately before it.
Mixing up α and (1 − α) in exponential smoothing.
Memorising the formula without understanding it.
Fix: α always multiplies the latest actual demand. (1 − α) multiplies the previous forecast.
Adding signed errors when calculating MAD.
Forgetting that positive and negative errors cancel.
Fix: Take the absolute value of each error before adding, then divide by the number of periods.
Recommending a quantitative method for a new product with no history.
Preferring calculations to judgement.
Fix: Where past data is absent, name qualitative methods such as Delphi, market survey or executive opinion.
Stating a forecast without a starting value for smoothing.
Missing the initial forecast in the question.
Fix: Use the forecast given. If none is given, state the assumption, such as taking the first actual demand as the first forecast.

Worked examples

Example 1

A Pune manufacturer recorded sales of a component over five months: April 400, May 440, June 420, July 460, August 480 units. (a) Forecast September using a 3-month simple moving average. (b) Forecast September using a 3-month weighted moving average with weights 0.5 (latest), 0.3 and 0.2.

Show the solution

(a) The last three months are June 420, July 460, August 480.
Sum = 420 + 460 + 480 = 1,360.
Forecast = 1,360 ÷ 3 = 453.33 units.
(b) August gets 0.5, July 0.3, June 0.2. The weights add to 1.
0.5 × 480 = 240; 0.3 × 460 = 138; 0.2 × 420 = 84.
Forecast = 240 + 138 + 84 = 462 units.

Answer: (a) About 453 units. (b) 462 units. The weighted forecast is higher because it gives more weight to the latest, higher sales.

Example 2

A firm's forecast for January was 100 units and actual demand was 110 units. Using exponential smoothing with α = 0.3, forecast February. If February actual is 120 units, forecast March. Also state the MAD of the January and February forecasts.

Show the solution

February forecast = 0.3 × 110 + 0.7 × 100 = 33 + 70 = 103 units.
March forecast = 0.3 × 120 + 0.7 × 103 = 36 + 72.1 = 108.1 units.
January error = 110 − 100 = 10.
February error = 120 − 103 = 17.
MAD = (10 + 17) ÷ 2 = 13.5 units.

Answer: February forecast 103 units; March forecast 108.1 units; MAD for January and February = 13.5 units.

Exam tips

Theory questions often ask for the difference between Delphi and market survey. Write at least four points in a table-like list: participants, anonymity, rounds, use.
In numerical questions, show the formula and each period's working. Step marks are given even if the final figure is wrong.
Keep the roles of qualitative and quantitative methods clear: judgement for new or uncertain situations, data for stable patterns.
In MCQs, check whether the question asks for the forecast or the error. The sign of the error can decide the option.
Mention limitations in long answers: moving average lags behind trends, and smoothing depends on the choice of α.

Practice questions from Operations Planning

Forecasting Demand for Operations: frequently asked questions

What is the difference between qualitative and quantitative forecasting methods?

Qualitative methods use opinion and judgement, such as Delphi, market survey and executive opinion. Quantitative methods use past data and mathematics, such as moving average, exponential smoothing and regression. Use qualitative methods when data is scarce and quantitative ones when reliable history exists.

What is the difference between the Delphi method and a market survey?

Delphi gathers forecasts from a panel of experts through several anonymous rounds with feedback until views converge. A market survey collects buying intentions from customers or potential customers, usually through questionnaires or interviews. Delphi relies on expert judgement, while a survey relies on customer response.

How does the value of α affect exponential smoothing?

A high α gives more weight to recent demand, so the forecast reacts quickly to changes. A low α gives more weight to the earlier forecast, so the forecast is smoother and reacts slowly. The choice depends on how stable demand is.

Which is better, simple or weighted moving average?

Neither is always better. A weighted moving average reacts faster when recent periods matter more. Compare the forecast error, such as MAD, on past data to decide.