Question

Discussion Question - There is an ongoing debate about the roles of quantitative and qualitative inputs in demand estimation and forecasting. Those in the qualitative camp argue that statistical analysis can only go so far. Demand estimates can be further improved by incorporating purely qualitative factors. Quantitative advocates insist that qualitative, intuitive, holistic approaches only serve to introduce errors, biases, and extraneous factors into the estimation task.

Suppose the executive for the theater chain is convinced that any number of bits of qualitative information (the identity of the director, the film’s terrific script and rock-music sound track, the Hollywood “buzz” about the film during production, even the easing of his ulcer) influence the film’s ultimate box-office revenue.

How might one test which approach—purely qualitative or statistical— provides better demand or revenue estimates? Are there ways to combine the two approaches? Provide concrete suggestions.

Answer 1

Quantitative techniques rely on the historical data and qualitative methods focusses on the informal methods basing on the experience and business experts judgement.

Quantitative demand forecast methods

Naive method: This method considers only the data of the last period and ignores the past data. The future forecast is equal to the last observed value in the data set.

Formula Ft = Yt-1 ,    Yt-1 is actual value in period t-1 , Ft  is forecast for period t

Moving average method: The historical data within a time period is considered to forecast the demand for upcoming period. The time perod is fixed and while calculating the demand for a new set of period, new data is considered within the fixed time period where the old data set is excluded. It calculates the trend and estimates the demand for a short term periods. It focusses on data values within limited time period and calculating the simple average.

Moving average =(n1+n2+...nN)N, where n is the data set per year, N is the number of year.

Exponential smoothing method: It considers all the data in the time series with larger weights added to recent value and lesser weights added to the distant data. This technique is useful if the data set in the time series does not display any trend or behavior.

Basic formula : St = ?yt-1+ (1 – ?) st-1

? = Smoothing constant, value ranges from 0 to 1,

t = time period

Data mining forecasting techniques: In recent years, there is huge volumes of data accumulated which makes it difficult for the forecasting experts to analyze such hige volumes in predicting the future trend or estimates. Data mining concept is useful in such scenarios which first selects the data basing on various time series models developed, data selected is tested and validated, statistical methods are adopted to choose the final model of forecasting and deployed using various technologies.

Qualitative methods:

Game theory : It studies the conflicts and cooperation between rational decision makers using various mathematical models. The interaction between the entities in the situation , within the expected outcome and set rules is estimated using game theory.

The delphi method: This method forecasts the demand using the expertize of the business groups reaching a consensual agreement with structures iterative models. It starts with selecting the experts, distributing the forecasting tasks, initial reports are submitted which are reviewed and feedback provided, final forecasts is developed basing on these reports in a consensual manner.