Page 1 of 5

Archives of Business Research – Vol. 9, No. 4

Publication Date: April 25, 2021

DOI:10.14738/abr.94.10022.

Kortam, W., & Soliman, A. (2021). Using Extrapolation Techniques of Big Data to Improve the Effectiveness of Sales Forecasting –

Empirical Evidence on the Egyptian Passengers Automobile Market. Archives of Business Research, 9(4). 91-95.

Services for Science and Education – United Kingdom

Using Extrapolation Techniques of Big Data to Improve the

Effectiveness of Sales Forecasting – Empirical Evidence on the

Egyptian Passengers Automobile Market

Wael Kortam

British University in Egypt

Amr Soliman

Cairo University Egypt

ABSTRACT

Sales forecasting is the tactical and strategic trigger for a significant number of

crucial functional and management activities. Extrapolation techniques will be used

as mathematical tools for demand forecasting, where extrapolation methods

estimates or generate values beyond the range of the original set of data. A

triangulated approach of experimental essences was applied to data analysis

composed of ARIMA, exponential smoothing and grey model GM(1,1). This research

is an endeavor to demonstrate the robustness of certain extrapolation models and

procedures with a view to word improving the effectiveness of sales forecast in

terms of relevance and accuracy. The forecasted sales for both brands is both

reserved and yet relatively optimistic. Thus, manifesting the high risk/high return

profile of the Egyptian passenger automobile market.

Key Words: Marketing, Sales Forecasting, Extrapolation.

INTRODUCTION

Sales forecasting is the tactical and strategic trigger for a significant number of crucial

functional and management activities. Such as demand management, capacity planning,

recruitments, investment analysis and risk management.

Sales forecasting model makers are intensely accumulating experience with advance statistical

modeling to sharpen their models, which are now readily user friendly for fast moving

consumer goods (FMCG), innovators and managers. Those models are quite less costly than

test markets and rollouts and enable diagnostic output as well as sensitivity testing.

Unfortunately when those models are adopted for forecasting more durable goods or services

they require massive amounts of data to work well and they are built heavily on assumptions

and are so complex that many decision makers find them too sophisticated. Actual managerial

practice within the context of durable products such as passenger automobile suggested that

using those models proved to be questionable as far as the accuracy and relevance of sales

forecast are concerns.

Extrapolation techniques will be used as mathematical tools for demand forecasting, where

extrapolation methods estimates or generate values beyond the range of the original set of data.

Page 2 of 5

92

Archives of Business Research (ABR) Vol. 9, Issue 4, April-2021

Services for Science and Education – United Kingdom

In this paper we will use Autoregressive Integrated Moving Average (ARIMA) and Exponential

Smoothing as parametric techniques and the Grey Model GM(1,1) as non-parametric technique.

ARIMA is an extension of the ARMA model when the time series data is non-stationary which

contain trend and seasonal pattern. This model is simply consists of three components, the

order of Autoregressive model, number of differences and the order of Moving Average model.

Exponential Smoothing is an important tool in most financial and engineering processes, as it

applies a continuous stabilization and adjustment to the model based on the past performance.

However, Grey model is a non-parametric mathematical tool used to forecast or generate data

with known and unknown characteristics.

LITERATURE REVIEW

It is surprising that most successful marketing research firm by far uses the simplest sales

forecasting methodology and requires the least data Purke Marketing Research which is a

division Nilsen adopts a portfolio of concept testing and product use testing and then calibrates

the trial and repeat percentages from their massive data files from previous studies.

Then the marketing research agency devises a set of experience – based heuristics (rules of

thumb) to translate those percentages to market shares. However, product managers and

innovators of durable products still most often use the simple versions of moving average and

time series and sometimes they don’t use any sales forecasting model at all and rely totally on

judgment based professional guess.

Intensive marketing and statistical research continuous toward improving the accuracy and

relevance of all sales forecasting models with special emphasis on contexts other than FMCG

markets.

Nevertheless, there are a number of challenges that make developing these new sales

forecasting models difficult and confusing. First, target customers don’t always know their

profiles, buying behaviors, needs and customer values regarding a certain product are. Even if

they do know such critical information for sales forecasting they may want to keep some

information from forecasters or offer false information. Complicating this problem is that

marketing research on those potential customers is often dangerously conducted. For example

there are so many Dracula effect horror stories about using focus groups for sales forecasting.

Second, sometimes competitors try to influence the accuracy and relevance of data collected

from resellers, regulators and market advisors.

Third, there is even lack of information on the internal management and support of durable

products inside the organization which delays official sales forecasting so that management

doesn’t commit themselves to obligations which might not be rendered.

Fourth, product managers are usually enthusiastic to get to the market which makes them push

for more optimistic yet unrealistic and irrelevant sales forecast.

Fifth, most common sales forecasting methodologies are extrapolations that work well on

established an stable patterns of data. Thus, they rely on mathematical and statistical

techniques that don’t necessarily reflect more underlying existing patterns and evidently

Page 3 of 5

93

Kortam, W., & Soliman, A. (2021). Using Extrapolation Techniques of Big Data to Improve the Effectiveness of Sales Forecasting – Empirical Evidence

on the Egyptian Passengers Automobile Market. Archives of Business Research, 9(4). 91-95.

URL: http://dx.doi.org/10.14738/abr.94.10022

emerging trends. Even forecasting methodologies that seems to be free of historical data and

stable assumptions like use of leaving indicators and causal models adopt relationships that has

been established in highly stable past.

RESEARCH GAP

The review of literature on sales forecasting from a marketing and statistical presptictves bin

points clearly to the need for further research on how to adopt extrapolation models to raise

the level of accuracy and relevance of sales forecasting especially within the context of durable

products.

AIM OF RESEARCH

This research is an endeavor to demonstrate the robustness of certain extrapolation models

and procedures with a view to word improving the effectiveness of sales forecast in terms of

relevance and accuracy. The Egyptian passenger automobile market is selected due to its need

for such more accurate and relevant forecast because of the globalization, deregulation and

customization pressures on marketers of major brands in the markets.

RESEARCH METHODOLOGY

This research will rely on a conclusive descriptive methodology based on longitudinal data

collection design of the sales figures of the two best sellers economy brands in the Egyptian

market of the past 2 decades (19 years). These two brands which are Hyundai and Kia (both

South Korean origin) were deliberately selected because of the relative size of the market

(around 50%) and their similarity in exposure to intensifying competition, differentiated

branding and drastic changes in customer bases on both national and global levels. These data

was collected by the major industry report produced by the Egyptian Association on Makers

and Importers of Cars (AMIC) which is an industry wide body founded to provide important

marketing and legal information on the current status and future outlook of the Egyptian

market for passenger and other kind of automobile.

A triangulated approach of experimental essences was applied to data analysis composed of

ARIMA, exponential smoothing and grey model GM(1,1). Those techniques was carefully

selected because it is argued by the authors to have more superior performance in improving

the accuracy and relevance of sales forecast within the context of passenger automobile as

durable products.

Obviously the limitation of the adopted methodology lie in relying on only two brands of only

new automobiles this regarding the magnitude of a huge Egyptian market of 13 countries of

origin, 83 major brands and 326 sub-brands alongside a considerable used cars market of

around 90,000 passenger cars per year.

MODELS

1. ARIMA(p,q)

Let X1

,X2 ... Xn denotes a set of times series and e1

,e2 ... endenotes a series of white noise

stochastic processes where et~ N(0, σ

2

).

Then Xt = μ + ∅1Xt−1 + ∅2Xt−2 + ⋯ + ∅pXt−p + et − θ1et−1 − θ2et−2 ... − θqet−q.