Home › Business Analysis › Forecasting
BDA

Decision Analysis: Forecasting

Business Decision Analysis — Study Guide

Ver: 0.1  ·  06 March 2020

⚠️   PAGE UNDER CONSTRUCTION
"The general who wins a battle makes many calculations beforehand the battle is fought. The general who loses a battle makes but few calculations beforehand. Thus do many calculations lead to victory, and few calculations to defeat: how much more no calculation at all! It is by attention to this point that I can foresee who is likely to win or lose."
— Sun Tzu (554–496 BC), Chinese philosopher & military general
§ 01

Forecasting

The notion of predicting all decisions involving the future, derived from relevant data. Forecasts aid decision making.

Surveys show statistical techniques are least used and qualitative techniques are more widely adopted — yet powerful computers are making statistical approaches easier than ever before.

Make forecasts by whatever means seem convenient and appropriate and with which decision makers feel comfortable, then compare forecast accuracy against actual outcomes. The best results often arise from combining multiple methods.

§ 02

Forecasting Steps

  1. Purpose
    Identify the purpose or objective of the forecast.
  2. Choice
    Choose which data needs to be forecast.
  3. Time
    Determine the time horizon of the forecast — e.g. one month or one year.
  4. Model
    Select the forecasting model(s) to be used.
  5. Acquire
    Acquire the data needed for the forecast.
  6. Validate
    Validate or test the model to determine its accuracy.
  7. Forecast
    Make the forecast.
  8. Implement
    Implement and act on the forecast.
§ 03

Forecasting Models

Model Type Description
Time Series Model Quantitative — Objective

Attempts to predict the future by using historical data. Based on a sequence of evenly spaced time periods (weeks, months, years). Has four components:

ComponentExplanation
TrendGradual upward or downward movement over time.
SeasonalityA pattern of demand fluctuation about the trend.
CyclesPatterns that occur every several years.
Random VariationCannot be accounted for by the three components above.

Sub-techniques include: Moving Average (weighting previous demand), Exponential Smoothing, and Trend Projections.
Causal Model Quantitative — Objective Examines the reason for changes in variables. For example, sales of ice cream are likely to vary according to weather — thus accurate weather prediction assists in forecasting ice cream sales.
Qualitative Model Descriptive — Subjective

Incorporates subjective factors. Contains the following sub-categories:

Long Range · Bias-Protected

Delphi Technique

Origin: Delphi town in 7th century BC Greece — home of the Oracle of Apollo. The Delphi technique is an iterative method of gaining forecasts from experts without requiring a meeting. A questionnaire is circulated; results are returned to experts who may revise views. Consensus is reached without distortion by dominant personalities.

Long Range · Alternative Futures

Scenario Writing

Constructs as many internally consistent, plausible, and significantly different scenarios of future events as possible. Science fiction authors have sometimes predicted future technologies such as atomic power, computers, and the internet.

Long Range · Form & Morphing

Morphological Analysis

Associated with Plato and Aristotle's study of form. A morphological matrix maps current technologies to forecast potential new combinations — e.g. a motor car = horse-drawn vehicle + steam engine. Leonardo Da Vinci imagined a helicopter five centuries ago. Helps identify technological opportunities and plan for disruption.

Simple

Executive Opinion

A small group of senior management convene to produce forecasts based on their expertise, sometimes in conjunction with statistical models.

Aggregated

Composite Method

Individual forecasts are aggregated and the aggregate is examined to determine if it is realistic. If not, individual forecasts need to be adjusted accordingly.

Survey-Based

Market Survey

Surveys existing or potential customers. Quantitative data is not the primary focus — e.g. people are surveyed about their TV habits and program preferences rather than being asked to supply viewing logs.
§ 04

Refresh: Concept of Probabilities

🔢 Objective Probability

Based on historic data & logic

  • Objectively recorded data has been used to establish the probability.
  • Consideration of the nature of the event can be logically determined.

🤔 Subjective Probability

Based on judgement or best guess

  • Not affected by the objective or subjective nature of the underlying probabilities.
  • Relies on personal judgement or best estimate.
§ 05

Concept of Events / States of Nature

Statistically Independent Events

When the occurrence of one event has absolutely no influence on the occurrence of the next event.

  • Getting a 6 on the first roll of a dice has no influence on subsequent rolls.
  • Probability of a Head (H) given a prior Tail (T): P(H|T) = P(H) — and similarly P(T|H) = P(T).
  • Statistical independence: conditional probability equals marginal probability.

Statistically Dependent Events

When the occurrence of one event does influence the probability of another event occurring.

  • A soccer team signing a world-class player improves their probability of winning the championship.
Bayes' Theorem 
-- Bayes' Theorem --
A mathematical approach for revising prior probabilities
by calculating conditional probabilities of statistically dependent events.

P(X|Y) = P(X ∩ Y) / P(Y)

Where:
  P(X|Y)     = probability of X given Y has occurred
  P(X ∩ Y)   = joint probability of both X and Y
  P(Y)       = marginal probability of Y