University of Birmingham · FACULTY OF MANAGEMENT

38004 Chap.11 Demand Forecasting Methods and Judgement

- one subject, every graph, every model, every mark
6 Chapters3-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 11 of 12 · 38004

Demand Forecasting Methods and Judgement

What forecasts are for

Weeks 8 and 9 treat demand forecasting as the input to decisions at three horizons: long-term capacity, medium-term workforce and resource plans, and short-term scheduling of staff, materials, inventory and maintenance.

The handbook stresses that forecasts are usually inaccurate, that error must be allowed for, that aggregate forecasts tend to be more accurate and that longer horizons are less reliable.

Methods that model data

Quantitative methods use past data.

For demand that holds near a stable level the module covers moving averages and exponential smoothing, compares them, and judges accuracy with error measures named MAD, MSD and MAPE.

The seminar adds that the method must suit the pattern, and that regression models predict without proving cause and effect.

Methods that rest on judgement

Qualitative methods suit new products and long horizons: economic indicators, market research, historical analogy, group methods, Delphi panels, sales force composites and scenario analysis.

Each has a weakness to manage, from costly research to biased estimates when forecasts feed sales targets. A good forecasting process combines a model baseline with judgement and records why any adjustment was made, so that later errors can be traced back to their cause.

In this chapter

What this chapter covers

  • 01

    Forecast horizons and the decisions they serve

  • 02

    Forecast error, aggregation and horizon length

  • 03

    Moving averages and exponential smoothing

  • 04

    Regression as prediction, not proof

  • 05

    Delphi panels and sales force composites

  • 06

    Separating forecasts from targets and preparation

Worked example · free

How fast 2-week and 4-week moving averages follow a demand step

Q [6 marks]. AskSia-authored practice. Weekly demand is 200 units in weeks 1 to 6 and 260 units from week 7, as in the moving average figure. Each forecast is the mean of the last 2 or the last 4 weeks. Find both forecasts for weeks 7 to 11 and the total absolute error over those weeks. Marks shown are AskSia's study weighting and not an official University allocation.
Figure 11.1.Figure 11.1. Weekly demand steps from 200 to 260 units in week 7; the 2-week moving average reaches the new level by week 9, while the 4-week moving average climbs through 215, 230 and 245 and only gets there in week 11.200260Week 1Week 12UnitsDemand2-week average4-week average
Figure 11.1. Weekly demand steps from 200 to 260 units in week 7; the 2-week moving average reaches the new level by week 9, while the 4-week moving average climbs through 215, 230 and 245 and only gets there in week 11.
  • 22-week forecasts: 200, 230, 260, 260, 260 for weeks 7 to 11; misses 60, 30, 0, 0, 0, total 90 units.
  • 24-week forecasts: 200, 215, 230, 245, 260; misses 60, 45, 30, 15, 0, total 150 units.
  • 2Conclude: the shorter window reaches the new level two weeks sooner and misses less after a lasting step.
After a lasting rise the 2-week average catches up by week 9 and the 4-week average only by week 11, so the short window misses 90 units in total against 150. The same speed would make it overreact to a one-off spike, so the choice of window depends on whether changes in this demand usually last.
Sia tip — Before shortening a moving-average window, check whether the latest jump is likely to persist.
Glossary

Key terms

Forecast Horizon
How far ahead a forecast looks; longer horizons serve capacity decisions but are less reliable.
Moving Average
A forecast equal to the mean demand of the most recent few periods, recalculated as each period ends.
Delphi Method
A structured process that gathers independent expert estimates over several rounds with summarised feedback.
Sales Force Composite
A forecast assembled from sales representatives' estimates of what their own customers will buy.
Leading Indicator
An economic series whose movements tend to signal a change in activity before it happens.
FAQ

Demand Forecasting Methods and Judgement FAQ

Why do managers use forecasts if they are usually wrong?

Capacity, staffing and purchasing decisions have to be made before demand is known, and a forecast with a known error range is better than an implicit guess. The handbook stresses planning for error: aggregate forecasts and shorter horizons are more reliable, so commitments that are hard to reverse deserve the most attention to uncertainty.

When should I use a short or a long moving-average window?

A short window reacts quickly, so it suits demand that genuinely shifts, but it also chases one-off spikes. A long window smooths noise but lags behind lasting changes. Exponential smoothing is the module's other method for stable demand and leans more on recent periods. Investigate the cause of an unusual week before deciding how strongly the forecast should respond.

Does a close regression fit prove that a factor drives demand?

No. Ice-cream sales and sunburn cases rise together because hot weather drives both, not because one causes the other. A regression in the module is a mathematical model of past data, so a close fit shows association only. Its forecasts still carry error, are more reliable for aggregated demand and lose accuracy as the horizon lengthens.

How does the Delphi method reduce bias in group forecasts?

Experts give estimates independently, so a senior or forceful voice cannot steer the room. The organiser summarises typical and extreme answers, asks those at the extremes to justify them, and repeats the rounds until a clear forecast emerges. It can be slow and costly, and choosing the panel size and number of rounds needs judgement.

Why can a sales force composite be biased?

Representatives know their customers well, which is valuable for complex product mixes and short horizons. But when the same forecast is used to set their targets, bias may creep in, often towards forecasting low, and detail about individual customers can be lost when figures are aggregated. Errors that always fall on one side are the warning sign.

Study strategy

Exam move

Start by listing the three decision horizons and writing beside each a decision you would base on a forecast, with the level of detail it needs. Then compute moving averages by hand for a short series that steps up, once with a two-period and once with a four-period window, and note how many periods each needs to catch up. Describe exponential smoothing and the three error measures in your own words.

For the judgemental side, make a table of the qualitative methods with one strength and one weakness each. For the Bass model, explain in two sentences how innovators and imitators shape adoption of a product with no sales history. Close by writing two sentences that explain why a sales team's forecast and its sales target should be set separately.

Working through Demand Forecasting Methods and Judgement in 38004? Sia is AskSia’s AI Management tutor — ask any 38004 Demand Forecasting Methods and Judgement question and get a clear, step-by-step explanation grounded in how 38004 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 17 of your subjects at University of Birmingham, plus 1,000+ more Bibles across the AskSia library.
Sia - your 38004 tutor, unlimited, worked the way the exam marks it
The full 3-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works
38004 · Organisational Decision Making and Operations Management - independent study guide on the AskSia Library. More University of Birmingham subjects · Microeconomics across all universities