Descriptive Analytics and Visualisation

This research paper explores the concepts of descriptive analytics and visualisation. It focuses on modelling quantity ordered, likelihood of recommendation, and forecasting production. The study aims to help businesses maintain a competitive advantage in the dynamic business environment. The research is conducted on Mad Dog Craft Beer, an Australian brewery company. The paper provides descriptive statistics, models, and forecasting techniques for the company's pale ale beer line. The study also emphasizes the importance of research, innovation, and technology in the modern business environment.

DESCRIPTIVE ANALYTICS AND VISUALISATION

DESCRIPTIVE ANALYTICS AND VISUALISATION
Contents
INTRODUCTION.......................................................................................................................................3
1. DEPENDENT VARIABLES DESCRIPTION....................................................................................4
1. MODELLING QUANTITY ORDERED............................................................................................6
2.1 FACTOR IDENTIFICATION...........................................................................................................6
2.2 MODEL BUILDING.........................................................................................................................6
2.3 INTERACTION EFFECT.................................................................................................................8
2. MODELLING LIKELIHOOD OF RECOMMENDATION................................................................9
3.1 MODEL.............................................................................................................................................9
3.2 PREDICTED PROBABILITIES.....................................................................................................11
3.3 PREDICTED PROBABILITIES VISUALIZATIONS....................................................................11
3. FORECASTING PRODUCTION.....................................................................................................11
CONCLUSION.........................................................................................................................................13
REFERENCES..........................................................................................................................................14
APPENDICES...........................................................................................................................................16
APPENDIX 1........................................................................................................................................16
APPENDIX 2........................................................................................................................................17
APPENDIX 3........................................................................................................................................22
APPENDIX 4........................................................................................................................................24
2

DESCRIPTIVE ANALYTICS AND VISUALISATION
INTRODUCTION
The dynamic nature of the modern business environment has made it necessary for businesses to
become dynamic as well (Kiechel, 2010; Sirajuddin, et al., 2017). Research, innovations and
technology have played a major role in making the modern business environment dynamic (Aalst
& Wil, 2011). Thus, for businesses to become dynamic, it is necessary to for them to invest in
research, innovations and technology. Among these three factors, research acts as a foundation
for the other two factors.
Through research, innovators get access to previously done work in the same field thereby
enabling them to develop better innovations (Albright & Winston, 2014). Also through research,
the optimal technology can be identified for a business process (Laudon & Guercio, 2014).
Optimal in this case implying a technology that is efficient will still being up to date. Hence
research can be considered as vital for acquiring and maintaining competitive advantage in the
modern dynamic business environment.
Mad Dog Craft Beer is an Australian based brewery company. This micro-brewery has managed
to establish a competitive advantage with fast growth (both in terms of sales and production) in
its customer bases in Regional Victoria and Melbourne. The company is interested in
maintaining its competitive advantage especially with the increase in competition from new
micro-brewery within its customer bases.
This research paper aims at conducting a research on the Order Quantity and Recommendations
of Mad Dog Craft Beer’s pale ale beer. Inferences can be drawn and actionable
recommendations determined from the analysis on the two variables. By using the inferences and
recommendations, Mad Dog Craft Beer can establish policies to enable it maintain its
3

DESCRIPTIVE ANALYTICS AND VISUALISATION
competitive advantage despite the rise in competition from new market entries. This paper will
also aim at developing a model for the forecasting of the production of Mad Dog Craft Beer’s
pale ale beer line. This forecasting model will enable the company to better plan their finances as
well as estimate the expected future profits.
The study paper will look at the descriptive statistics of the Order Quantity and Recommendation
variables, develop models for estimation of these two variables and develop a forecasting model
for the production of pale ale beer line.
1. DEPENDENT VARIABLES DESCRIPTION
The table, Table 1: Dependent Variable Summary Description, below gives the summary
description of the two dependent variables in this research study:
Table 1: Dependent Variable Summary Description
Variable Variable
Description
Nature of
Variable
Variable Type Measurement
Scale
Order_Qty Informs on the
number of pale
ale beers
bottles ordered
by a customer.
Numerical
Variable.
Dependent
Variable.
Ratio Scale.
4

DESCRIPTIVE ANALYTICS AND VISUALISATION
Recommendation Informs on
whether a
customer
would
recommend the
pale ale beer to
others.
Categorical
Variable with
two levels; 0
represents No
and 1
represents Yes.
Dependent
Variable.
Nominal Scale.
From Table 3: Descriptive Statistics of Order_Qty and Recommend, we observe that
customers that purchase the pale ale beer are slightly more likely to recommend it to other
people than they are likely not to. 101 customers replied they would recommend the pale ale
beer to other people while 99 said they would not recommend it. The average order quantity
of the pale ale beer was 1.665 with a standard deviation of 0.8932 and mode of 7.2. The
lower quartile (Q1), median, Upper Quartile and Interquartile Range values for Order_Qty
were 7.1, 7.6, 8.2 and 1.1 respectively. The range for the Order_Qty data is 5.6 units. From
the histogram of the Order_Qty, Figure 1: Histogram of Order_Qty, we observe that the data
distribution can be described as skewed to the right (negatively skewed). This is also
confirmed from the value of Skewness = -0.2064 with a kurtosis of 0.5840.
5

DESCRIPTIVE ANALYTICS AND VISUALISATION
1. MODELLING QUANTITY ORDERED
2.1 FACTOR IDENTIFICATION
In order to identify the possible factors that affect the quantity ordered variable, Order_Qty,
the Factor Analysis Method was applied. Factor Analysis is a multivariate analysis method
that is used in dimension reduction to obtain a reasonably acceptable number of variables for
analysis (Freedman, 2009; Vicenc, 2017). The Principal Analysis method reduces the size of
a set of independent variables to a small manageable size for the analysis of the impact on the
dependent variable (Han & Jaiwei, 2011; O'Neil & Schutt, 2013).
From Table 4: Eigen Values, we observe that the following factors had an Eigen value
greater than 1; Loyalty (3.810), Cust_Type (2.764), Region (1.722) and Dist_Channel
(1.081). Therefore, we consider the Loyalty, Cust_Type, Region and Dist_Channel as the
possible factors affecting the quantity ordered of the pale ale beer line of the Mad Dog Craft
company. This can also be seen from Figure 2: Scree Plot, where there is a point of inflection
at Eigen value 5, thus we consider factor 1 to 4.
2.2 MODEL BUILDING
The purpose of this model is to estimate the quantity ordered for the pale ale beer line of the
Mad Dog Craft Beer company. The model will consider the possible factors that affect the
Order_Qty variable in modelling the estimation of the quantity ordered for the pale ale beer
line.
The factors that are considered are the Loyalty, Cust_Type, Region and Dist_Channel. These
factors will form the independent variables while the Order_Qty forms the dependent
6

DESCRIPTIVE ANALYTICS AND VISUALISATION
variable. The sample is composed of 200 observations of customers on the 5 variables of
interest for the model.
Regression analysis is used in the generation of the model for the estimation of the quantity
ordered for the pale ale beer line. Regression analysis is a statistical analysis technique that
estimates the value of one variable(s), known as the dependent variable(s), using another
variable(s), known as independent variable(s) (Jaulin, 2010). This study specifically applies
the multiple regression analysis technique for estimation. The multiple regression analysis is
a type of regression analysis whereby there are multiple independent variables estimating a
single dependent variable (Jorge, et al., 2013; Witten, 2011).
The first model contained all four independent variables with the summary output given in
Table 5: Model 1. The model had an AIC = -98.88 with the model equation given as below:
Orde rQty =6.866+ 0.0734 Loyalty+0.0182Cus tType −0.1214 Region+ 0.5499 Dis t Channel
The second model contained the following independent variables; Loyalty, Cust_Type and
Region. The summary output is given in Table 6: Model 2. The model had an AIC = -79.90
with the model equation given as below:
Orde rQty =7.120+0.0890 Loyalty−0.0570 Cus tType−0.1172 Region
The third model contained the following independent variables; Loyalty and Cust_Type with
the summary output given in Table 7: Model 3. The model had an AIC = -76.21 with the
model equation as given below:
Orde rQty =6.866+ 0.0867 Loyalty−0.0496 CustType
The fourth model had the following independent variables; Cust_Type, Region and
Dist_Type with the summary output given in Table 8: Model 4. The model had an AIC = -
73.12 with the model equation as given below:
7

DESCRIPTIVE ANALYTICS AND VISUALISATION
Orde rQty =7.355+0.0120 Cus tType−0.0353 Region+0.7076 Dis tChannel
The fifth model had the following independent variables: Region and Dist_Type with the
summary output given in Table 9: Model 5. The model had an AIC = -75.11 with model
equation as given below:
Orde rQty =7.362−0.0362 Region+ 0.7057 Dis tChannel
The sixth model had the following independent variables; Loyalty and Dist_Channel with the
summary output given in Table 10: Model 6. The model had an AIC = -101.7 with the model
equation as given below:
Orde rQty =6.805+ 0.0714 Loyalty+0.5822 Dis tChannel
The sixth model had the lowest AIC = -101.7, this implies that the sixth model was the best
fit model of the six iterations of the estimation model for the quantity ordered of the pale ale
beer line. The test for this model is given in Table 11: Model 6 Prediction. From this table we
observe that the model has a reasonably efficient prediction ability.
2.3 INTERACTION EFFECT
This model contains the following independent variables; Quality and Brand_Image with the
summary output given in Table 12: Model 7. The model had an AIC = 123.5 with the model
equation as given below:
Orde rQty =3.588+ 0.3094 Quality+0.3116 Bran dImage
From the model equation above, a unit change in the quality of the pale ale beer results in a
0.3094 change in the quantity perception of pale ale beer ordered. Also, a unit change in
8

DESCRIPTIVE ANALYTICS AND VISUALISATION
brand image perception of the pale ale beer results in a 0.3116 change in the quantity of pale
ale beer ordered. Considering an α = 0.05 level of significance, the p-values of both the
Quality and Brand_Image, 2.61e-14 and 1.54e-10 respectively, are lower than α = 0.05 level
of significance. Therefore, both the independent variables are significant predictors in the
estimation of the Order_Qty variable.
The value of the Adjusted R Squared = 0.3340, this implies that the model explains only
33.40% of the relationship between the variable in the model. Hence, despite having a very
low AIC = 123.5, the low value of the Adjusted R Squared indicates that the model is a poor
fit.
The regression analysis indicate that the quality perception and brand image perceptions are
significant predictors of the quantity of the order of pale ale beer. Since quality perception is
determined by the brand image perception, investing in the brand image for the pale ale beer
line would significantly improve the quantity of the orders of the pale ale beer.
2. MODELLING LIKELIHOOD OF RECOMMENDATION
3.1 MODEL
This modelling process utilizes logistic regression analysis to develop a predictive model for
the estimation of the likelihood that a customer will recommend pale ale beer to other people.
The logistic regression is a regression analysis technique that estimates a dichotomous
categorical response (dependent) variable using independent variables that are both
numerical and categorical (Barbara & Susan, 2014; Hosmer, 2013).
9

DESCRIPTIVE ANALYTICS AND VISUALISATION
The model has Recommendation as the dependent variable and the independent variables are;
Distribution Channel, Quality, Brand Image and Shipping Speed. The summary model output
is given in Table 13: Model 8 (Coefficients). The resultant model equation is as given below:
Recommendation=−13.28+ 0.9684 Dis tChannel +0.6541 Quality+0.6214 Bran dImage + 1.159 Shippin gSpeed
From the model we observe that a change in the distribution channel results in a 0.9684
change in the likelihood of recommendation, a unit change in the quality perception results in
a 0.6541 change in the likelihood of recommendation, a unit change in the brand image
perception results in a 0.6214 change in the likelihood of recommendation and a unit change
in the shipping speed results in a 1.159 change in the likelihood of recommendation.
The p-values of the variables are 0.0102, 2.1e-05, 0.0015 and 5.7e-05 for Distribution
Channel, Quality, Brand Image and Shipping Speed respectively. Considering an α = 0.05
level of significance, then all the p-values are lower than α = 0.05 level of significance and
hence all the variables are significant predictors of the likelihood of a customer
recommending the pale ale beer to other people.
From the chi-square test in Table 14: Model 8 (Tests), the p-value = 7.35e-18, which is less
than α = 0.05 level of significance. Thus, we conclude that the model is fit and significant for
the estimation of the likelihood of a customer recommending a pale ale beer to other people.
3.2 PREDICTED PROBABILITIES
Table 15:Predicted Probabilities for Model 8 considers a case for 20 customers that
purchased the pale ale beer from Mad Dog Craft Beer company. From this table we observe
that as the quality perception and the brand image perception increase, the likelihood of a
10

DESCRIPTIVE ANALYTICS AND VISUALISATION
customer recommending the pale ale beer to other people equally increases. This is true for
both the customers who purchase pale ale beer from sales representative and directly from
the Mad Dog Craft Beer company. Also in both cases, customers that consider the brand
image as poor or as neutral, are less likely to recommend pale ale beer to other people.
3.3 PREDICTED PROBABILITIES VISUALIZATIONS
The plots Figure 3: Predicted Probability of Recommendation for Customers Purchasing
from Sales Representative and Figure 4: Predicted Probability of Recommendation for
Customers that Purchase Directly respectively show the graphs of the predicted probabilities
of recommendation for customers that purchase from sales representatives and customers that
purchase directly from Mad Dog Craft Beer company. The two plots generally have the same
trend and indicate that as the quality perception and the brand image perception increase, the
likelihood of a customer recommending the pale ale beer to other people equally increases.
3. FORECASTING PRODUCTION
The SARIMA model was applied in the forecasting of the production of pale ale beer. The
SARIMA is a special case of an ARIMA model that is applied in instances when there is an
aspect on seasonality in the time series data (Galit, et al., 2018; Shaffer, 2011). The data on
the production of pale ale beer is recorded quarterly, hence introducing an aspect of
seasonality in the time series data for the production. Thus, the SARIMA model is
appropriate for the forecasting of the production of the pale ale beer since it caters for the
seasonality present in the data.
11

DESCRIPTIVE ANALYTICS AND VISUALISATION
The summary output of the model is given in Table 16: SARIMA Model. From the summary
output, we observe that the model for forecasting the production of pale ale beer can be given
as follows:
Productio nt =−1.703+0.8779 Productio nt−1−0.0452 Productio ns−1 + e−1.342 et−1−0.6243 es −1
From the model above, we observe that future production has a 0.8779 per unit dependence
on the production of the previous quarter and a -0.0452 per unit dependence on the
production of the previous season (year). The error due to the previous quarter accounts for -
1.342 per unit while the error due to the previous season (year) accounts for -0.6243 per unit.
The SARIMA model forecast for the turnover in production of the pale ale beer line for the
next four quarters as given in Table 2: Forecast Turnover in Production of Pale Ale Beer Line
below:
Table 2: Forecast Turnover in Production of Pale Ale Beer Line
Year Quarter Forecasted Turnover in
Production
2019 Q2 1728.22
Q3 2039.05
Q4 1608.08
2020 Q1 1722.63
CONCLUSION
From the analysis done and the inferences drawn in this research study we can conclude that the
main factors that significantly affect the quantity of order of the pale ale beer line are Loyalty,
Distribution Channel, Quality and Brand Image. However, the two models that have these
12

DESCRIPTIVE ANALYTICS AND VISUALISATION
variables as the independent variables are not fit and hence cannot be considered as efficient in
the estimation of the quantity of order of the pale ale beer line.
The model for the estimation of the likelihood of a customer recommending the pale ale beer to
other people has the following independent variables; Distribution Channel, Quality, Brand
Image and Shipping Speed. This model is fit for the estimation of the likelihood of
recommendation. A predictive analysis using this model revels that as the quality perception and
the brand image perception increase, the likelihood of a customer recommending the pale ale
beer to other people equally increases.
The forecasting model for the turnover in production shows that the production will continue
rising through to the third quarter of 2019, but then drops in the final quarter of the year before
rising again during the first quarter of the year 2020.
REFERENCES
Aalst, VD & Wil, MP 2011, Process Mining: Discovery, Conformance and Enhancement of
Business Processes, 1st edn, Springer, New York.
13

DESCRIPTIVE ANALYTICS AND VISUALISATION
Albright, CS & Winston, WL 2014, Business Analytics: Data Analysis & Decision Making. 2nd
edn, Cengage Learning, New York.
Barbara, I & Susan, D 2014. Introductory Statistics. 1st edn, OpenStax CNX, New York.
Freedman, DA 2009, Statistical Models: Theory and Practice. 1st edn, Cambridge University
Press, London.
Galit, S. et al. 2018, Data Mining for Business Analytics, 1st edn, John Wiley & Sons, Inc., New
Delhi.
Han, K & Jaiwei, P 2011, Data Mining: Concepts and Techniques, 3rd edn, Morgan Kaufman,
London.
Hosmer, D 2013, Applied Logistic Regression, 1 edn, Wiley, Hoboken, New Jersey.
Jaulin, L 2010, "Probabilistic set-membership approach for robust regression", Journal of
Statistical Theory and Practice, Vol.5, No.1, pp. 1-14.
Jorge, AA, Angela, A & Edson, ZM 2013, "Robust Linear Regression Models: Use of Stable
Distribution for the Response Data", Open Journal of Statistics, Vol.3, No.7, pp. 3-5.
Kiechel, W 2010, The Lords of Strategy, 2nd edn, Havard Business Press, New York.
Laudon, KC & Guercio, TC 2014, E-commerce. Business. Technology. Society, 1st edn, Pearson,
Chicago.
O'Neil, C & Schutt, R 2013, Doing Data Science, 3rd edn, O'Reily, London.
Shaffer, CA 2011, Data Structures and Algorithms Analysis, 1st edn, Dover, Mineola.
14

DESCRIPTIVE ANALYTICS AND VISUALISATION
Sirajuddin, O, Ibrahim, AH & Jamali, H 2017, "Five Competitive Forces Model and the
Implementation of Porter's Generic Strategies to Gain Firm Performances", Science Journal of
Business and Management, Vol.10, No.16, pp. 1-5.
Vicenc, T 2017, Studies in Big Data. 1st edn, Springer International Publishing, Chicago.
Witten, IH 2011, Data Mining: Practical Machine Learning Tools, 3rd edn, Morgan Kaufmann,
Sydney.
15

DESCRIPTIVE ANALYTICS AND VISUALISATION
APPENDICES
APPENDIX 1
Table 3: Descriptive Statistics of Order_Qty and Recommend
Order_Qty
Mean 7.665
Median 7.6
Mode 7.2
Standard
Deviation
0.893232513
Range 5.6
Q1 7.1
Q3 8.2
IQR 1.1
Kurtosis 0.584037705
Skewness -0.206346633
Recommend
0 99
1 101
Figure 1: Histogram of Order_Qty
16

DESCRIPTIVE ANALYTICS AND VISUALISATION
APPENDIX 2
Table 4: Eigen Values
Variable Eigen
Value
Loyalty 3.80975
9
Cust_Type 2.76345
8
Region 1.72232
8
Dist_Channel 1.08104
5
Quality 0.87592
SM_Presence 0.62387
2
Advert 0.51752
2
Brand_Image 0.40757
8
Comp_Pricing 0.36699
1
Order_Fulfillment 0.30185
8
Flex_Price 0.27397
5
Shipping_Speed 0.14739
6
Shipping_Cost 0.10829
8
17

DESCRIPTIVE ANALYTICS AND VISUALISATION
1 2 3 4 5 6 7 8 9 10 11 12 13
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
Scree Plot
Eigen Value
%
Figure 2: Scree Plot
Table 5: Model 1
18

DESCRIPTIVE ANALYTICS AND VISUALISATION
Table 6: Model 2
Table 7: Model 3
19

DESCRIPTIVE ANALYTICS AND VISUALISATION
Table 8: Model 4
Table 9: Model 5
20

DESCRIPTIVE ANALYTICS AND VISUALISATION
Table 10: Model 6
Table 11: Model 6 Prediction
Actual Value Predicted
Value
8.4 7.6014
7.5 7.7332
9.0 8.244
7.2 7.8046
9.0 8.0298
6.1 7.3762
7.2 6.9478
7.7 7.4476
8.2 7.3048
6.7 7.4476
21

DESCRIPTIVE ANALYTICS AND VISUALISATION
Table 12: Model 7
APPENDIX 3
Table 13: Model 8 (Coefficients)
22

DESCRIPTIVE ANALYTICS AND VISUALISATION
Table 14: Model 8 (Tests)
Table 15:Predicted Probabilities for Model 8
23

DESCRIPTIVE ANALYTICS AND VISUALISATION
1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Predicted Probability of Recommendation for
Customers Purchasing from Sales Representative
Quality and Brand Image
Predicted Probability of Recommendation
Figure 3: Predicted Probability of Recommendation for Customers Purchasing from Sales Representative
1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Predicted Probability of Recommendation for
Customers that Purchase Directly
Quality and Brand Image
Predicted Probability of Recommendation
Figure 4: Predicted Probability of Recommendation for Customers that Purchase Directly
24

DESCRIPTIVE ANALYTICS AND VISUALISATION
APPENDIX 4
Table 16: SARIMA Model
25