PDF Recommender System For Improving Customer Loyalty Katarzyna Tarnowska Ebook Full Chapter

Recommender System for Improving
Customer Loyalty Katarzyna Tarnowska

Visit to download the full and correct content document:
https://textbookfull.com/product/recommender-system-for-improving-customer-loyalty-
katarzyna-tarnowska/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Customer Loyalty and Supply Chain Management Business

to Business Customer Loyalty Analysis 1st Edition Ivan
Russo
https://textbookfull.com/product/customer-loyalty-and-supply-
chain-management-business-to-business-customer-loyalty-
analysis-1st-edition-ivan-russo/
Marketing and Customer Loyalty The Extra Step Approach

1st Edition Mauro Cavallone (Auth.)
https://textbookfull.com/product/marketing-and-customer-loyalty-
the-extra-step-approach-1st-edition-mauro-cavallone-auth/
The Opt Out Effect Marketing Strategies that Empower

Consumers and Win Customer Driven Brand Loyalty 1st
Edition Gerald E. Smith
https://textbookfull.com/product/the-opt-out-effect-marketing-
strategies-that-empower-consumers-and-win-customer-driven-brand-
loyalty-1st-edition-gerald-e-smith/
Fashion Recommender Systems Nima Dokoohaki
https://textbookfull.com/product/fashion-recommender-systems-
nima-dokoohaki/
Practical Recommender Systems 1st Edition Kim Falk
https://textbookfull.com/product/practical-recommender-
systems-1st-edition-kim-falk/
Customer Accounting: Creating Value with Customer

Analytics Massimiliano Bonacchi
https://textbookfull.com/product/customer-accounting-creating-
value-with-customer-analytics-massimiliano-bonacchi/
Managing Customer Relationships Using Customer Care

Techniques 1st Edition Brzozowska
https://textbookfull.com/product/managing-customer-relationships-
using-customer-care-techniques-1st-edition-brzozowska/
Recommender System with Machine Learning and Artificial

Intelligence Practical Tools and Applications in
Medical Agricultural and Other Industries 1st Edition
Sachi Nandan Mohanty (Editor)
https://textbookfull.com/product/recommender-system-with-machine-
learning-and-artificial-intelligence-practical-tools-and-
applications-in-medical-agricultural-and-other-industries-1st-
edition-sachi-nandan-mohanty-editor/
Group Recommender Systems An Introduction 1st Edition

Alexander Felfernig
https://textbookfull.com/product/group-recommender-systems-an-
introduction-1st-edition-alexander-felfernig/
Studies in Big Data 55
Katarzyna Tarnowska
Zbigniew W. Ras
Lynn Daniel
Recommender
System for
Improving
Customer Loyalty
Studies in Big Data
Volume 55
Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Big Data” (SBD) publishes new developments and advances
in the various areas of Big Data- quickly and with a high quality. The intent is to
cover the theory, research, development, and applications of Big Data, as embedded
in the fields of engineering, computer science, physics, economics and life sciences.
The books of the series refer to the analysis and understanding of large, complex,
and/or distributed data sets generated from recent digital sources coming from
sensors or other physical instruments as well as simulations, crowd sourcing, social
networks or other internet transactions, such as emails or video click streams and
others. The series contains monographs, lecture notes and edited volumes in Big
Data spanning the areas of computational intelligence including neural networks,
evolutionary computation, soft computing, fuzzy systems, as well as artificial
intelligence, data mining, modern statistics and Operations research, as well as
self-organizing systems. Of particular value to both the contributors and the
readership are the short publication timeframe and the world-wide distribution,
which enable both wide and rapid dissemination of research output.
** Indexing: The books of this series are submitted to ISI Web of Science, DBLP,
Ulrichs, MathSciNet, Current Mathematical Publications, Mathematical Reviews,
Zentralblatt Math: MetaPress and Springerlink.
More information about this series at http://www.springer.com/series/11970

Katarzyna Tarnowska Zbigniew W. Ras
• •
Lynn Daniel
Recommender System
for Improving Customer
Loyalty
123
Katarzyna Tarnowska Zbigniew W. Ras
Department of Computer Science Department of Computer Science
San Jose State University University of North Carolina
San Jose, CA, USA Charlotte, NC, USA
Polish-Japanese Academy
Lynn Daniel
of Information Technology
The Daniel Group
Warsaw, Poland
Charlotte, NC, USA
ISSN 2197-6503 ISSN 2197-6511 (electronic)

Studies in Big Data
ISBN 978-3-030-13437-2 ISBN 978-3-030-13438-9 (eBook)
https://doi.org/10.1007/978-3-030-13438-9
Library of Congress Control Number: 2019932683
© Springer Nature Switzerland AG 2020

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This book presents a novel data-driven approach to solve the problem of improving
customer loyalty and customer retention. The data mining concepts of action rules
and meta-actions are used to extract actionable knowledge from customer survey
data and build a knowledge-based recommender system (CLIRS—Customer
Loyalty Improvement Recommender System). Also, a novel approach to extract
meta-actions from the text is presented. So far, the use of meta-actions required a
pre-defined knowledge of the domain (i.e., medicine). In this research, an automatic
extraction of meta-actions is proposed and implemented by applying Natural
Language Processing and Sentiment Analysis techniques on the customer reviews.
The system’s recommendations were optimized by means of implemented mech-
anism of triggering optimal sets of action rules. The optimality of recommendations
was defined as maximal Net Promoter Score impact given minimal changes in the
company’s service. Also, data visualization techniques are proposed and imple-
mented to improve understanding of the multidimensional data analysis, data
mining results, and interacting with the recommender system’s results.
Another important contribution of this research lies in proposing a strategy for
building a new set of action rules from text data based on sentiment analysis and
folksonomy. This new approach proposes a strategy for building recommendations
directly from action rules, without triggering them by meta-actions. The coverage
and accuracy of the opinion mining were significantly improved within a series of
experiments, which resulted in better recommendations. Therefore, the research
presents a novel approach to build a knowledge-based recommender system
whenever only text data is available.
San Jose, USA Katarzyna Tarnowska

Charlotte, USA/Warsaw, Poland Zbigniew W. Ras
Charlotte, USA Lynn Daniel
v
About the Book
In this book, data mining and text mining techniques are a proposed approach to the
problem of customer loyalty improvement. The built solution is an automated
data-driven and user-friendly recommender system based on actionable knowledge
and sentiment analysis.
The system proved to work in real settings and its results have already been
discussed with the end business users. Its main value lies in suggesting and
quantifying the effectiveness of the course of strategic actions to improve the
company’s growth potential. Another strength of the approach is that it works on
the overall knowledge in the industry, which means that worse-performing com-
panies can learn from the knowledge and experience of their better-performing
competition.
In this book, we introduce the problem area and describe the dataset on which
the work has been done. We present the background knowledge and techniques
necessary to understand the built solution, as well as the current state of the art
applications in the researched area.
Further, we describe all the work that already has been done within this research
project, including: visual techniques applied to enhance interactiveness and
friendliness of the system, experiments on improving the knowledge miner and
finally the new architecture and the implementation of the system that works solely
on the text customer feedback comments. Lastly, we focus our research on finding
and testing new ways of improving the algorithm for natural language processing of
text comments and we present the results.
Within the work done, we identified new topics that need further research and
improvement, which will become the focus towards the future directions in this
research.
vii
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........ 1
1.1 Why Customer Experience Matters More Now? . . . . ........ 1
1.2 Top (and Bottom) Line Reasons for Better Customer
Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........ 3
1.3 What is Next? . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........ 5
1.4 Final Observations . . . . . . . . . . . . . . . . . . . . . . . . . ........ 5
2 Customer Loyalty Improvement . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Introduction to the Problem Area . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Dataset Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Decision Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Problem Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4.1 Attribute Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4.2 Attribute Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.3 Customer Satisfaction Analysis and Recognition . . . . . . 10
2.4.4 Providing Recommendations . . . . . . . . . . . . . . . . . . . . . 11
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1 Customer Satisfaction Software Tools . . . . . . . . . . . . . . . . . . . 13
3.2 Customer Relationship Management Systems . . . . . . . . . . . . . . 14
3.3 Decision Support Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Recommender Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.1 Recommender Systems for B2B . . . . . . . . . . . . . . . . . . 15
3.4.2 Types of Recommender Systems . . . . . . . . . . . . . . . . . . 16
3.4.3 Knowledge Based Approach for Recommendation . . . . . 17
3.5 Text Analytics and Sentiment Analysis Tools . . . . . . . . . . . . . . 17
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
ix
x Contents
4 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1 Knowledge Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.1 Decision Reducts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.2 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.3 Action Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.4 Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Text Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.1 Sentiment Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.2 Aspect-Based Sentiment Analysis . . . . . . . . . . . . . . . . . 27
4.2.3 Aspect Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2.4 Polarity Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2.5 Natural Language Processing Issues . . . . . . . . . . . . . . . 32
4.2.6 Summary Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.7 Visualizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.8 Measuring the Economic Impact of Sentiment . . . . . . . . 35
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5 Overview of Recommender System Engine . . . . . . . . . . . . . . . . . . . 41
5.1 High-Level Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Data Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2.1 Raw Data Import . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2.2 Data Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.3 Semantic Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.4 Hierarchical Agglomerative Method for Improving NPS . . . . . . 51
5.5 Action Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.6 Meta Actions and Triggering Mechanism . . . . . . . . . . . . . . . . . 54
5.7 Text Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6 Visual Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 59
6.1 Decision Reducts as Heatmap . . . . . . . . . . . . . . . . . . . . . . ... 59
6.2 Classification Visualizations: Dual Scale Bar Chart
and Confusion Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3 Multiple Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.4 Evaluation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.4.1 Single Client Data (Local) Analysis . . . . . . . . . . . . . . . 63
6.4.2 Global Customer Sentiment Analysis and Prediction . . . 64
6.5 User-Friendly Interface for the Recommender System . . . . . . . . 65
7 Improving Performance of Knowledge Miner . . . . . . . . . . . . . . . . . 69
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.3 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
7.4 Strategy and Overall Approach . . . . . . . . . . . . . . . . . . . . . . . . 71
Contents xi
7.5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.5.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.5.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.5.3 New Rule Format in RS . . . . . . . . . . . . . . . . . . . . . . . . 78
7.6 Plans for Remaining Challenges . . . . . . . . . . . . . . . . . . . . . . . 85
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8 Recommender System Based on Unstructured Data . . . . . . . . . . . . 87
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
8.3 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
8.4.1 Data Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.4.2 Action Rule Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
8.4.3 Ideas for the Improvement of Opinion Mining . . . . . . . . 91
8.4.4 Sentiment Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8.4.5 Polarity Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
8.5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.5.1 Initial Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.5.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.5.3 Improving Sentiment Analysis Algorithm . . . . . . . . . . . 94
8.5.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . 98
8.5.5 Modified Algorithm for Opinion Mining . . . . . . . . . . . . 100
8.5.6 Comparing Recommendations with the Previous
Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.6.1 Complex and Comparative Sentences . . . . . . . . . . . . . . 107
8.6.2 Implicit Opinions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.6.3 Feature and Opinion in One Word . . . . . . . . . . . . . . . . 108
8.6.4 Opinion Words in Different Context . . . . . . . . . . . . . . . 109
8.6.5 Ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.6.6 Misspellings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.6.7 Phrases, Idiomatic and Phrasal Verbs Expressions . . . . . 110
8.6.8 Entity Recognition From Pronouns and Names . . . . . . . 110
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
9 Customer Attrition Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
9.3 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9.4.1 Automatic Data Labelling . . . . . . . . . . . . . . . . . . . . . . . 115
9.4.2 Pattern Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
xii Contents
9.4.3 Sequence Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

9.4.4 Action Rule, Meta Action Mining and Triggering . . . . . 117
9.5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
9.5.1 Initial Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
9.5.2 Attribute Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
9.5.3 Classification Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
9.5.4 Action Rule Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
10.1 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
10.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
List of Figures
Fig. 2.1 The concept of Net Promoter Score as a way to quantify

and categorize customer satisfaction . . . . . . . . . . . . . . . . . . . . .. 8
Fig. 2.2 Illustration of the NPS dataset structure—the features
and the decision attribute . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 8
Fig. 4.1 Feature-based opinion summarization process. Source [15] . . .. 28
Fig. 4.2 Architecture of the system for mining Feature-Opinion
words based on syntactical dependency. Source [22] . . . . . . . .. 30
Fig. 4.3 Example of syntactical dependency tree. Source [22] . . . . . . . .. 30
Fig. 4.4 Example of summary generation. Source [15] . . . . . . . . . . . . .. 33
Fig. 4.5 Multivariate comparison across scale and geography
showing correlation in [35] . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 34
Fig. 4.6 Visualization for aspect-based review summarization
developed within [28] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 34
Fig. 4.7 Rose plots technique for visualizing an affective content
of the documents. Source [36] . . . . . . . . . . . . . . . . . . . . . . . . .. 35
Fig. 4.8 Principal-Components-analysis visualization of associations
between products (squares) and automatically selected
opinion-oriented terms (circles). Source [37] . . . . . . . . . . . . . .. 36
Fig. 5.1 Architecture and main use-cases of the data-driven
and user-friendly customer loyalty improvement recommender
system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 42
Fig. 5.2 Control flow task definition for the process of importing
the data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 44
Fig. 5.3 Data flow task definition for the process of exporting data
from Excel to database in two vertical partitions (division
based on columns) and adding Index column . . . . . . . . . . . . . .. 45
Fig. 5.4 Data flow task definition for the process of merging rows
based on Index column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 46
Fig. 5.5 Data flow task definition for the process of combining two
parts of 2016 into one final table (merging data horizontally
with UNION ALL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 46
xiii
xiv List of Figures
Fig. 5.6 Initial manual preprocessing of the data before loading

it into recommender system . . . . . . . . . . . . . . . . . . . . . . . . . . .. 47
Fig. 5.7 Diagram showing automation of the data preprocessing
that can be also integrated into recommender system . . . . . . . .. 47
Fig. 5.8 Preprocessing steps implemented in MS SQL Server
Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 48
Fig. 5.9 Javascript-based visualization of the dendrogram showing
semantic similarity between clients in 2015: chosen Client9
with highlighted semantically similar clients ordered
by numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 51
Fig. 5.10 Results of running HAMIS procedure on 38 datasets
representing clients for Service survey data from 2016:
the number of clients by which a client was extended
and its original NPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 53
Fig. 6.1 Dynamic visualization of a reduct matrix based on a heatmap
design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 60
Fig. 6.2 Reduct heatmap with NPS row chart and NPS category
distribution chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 61
Fig. 6.3 Visualizations for the classifier’s results—accuracy, coverage
and confusion matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 61
Fig. 6.4 Visual multiple analysis for a single client (“Client1”) . . . . . . .. 63
Fig. 6.5 Classification results analysis—confusion matrices . . . . . . . . . .. 64
Fig. 6.6 Javascript-based visualization for depicting clients’ locations
and their semantic neighbors. Also, serves as an interface
for further analysis of a chosen entity . . . . . . . . . . . . . . . . . . . .. 66
Fig. 6.7 Javascript-based interactive visualization for exploring
recommendations options and their attractiveness based
on chosen feasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 66
Fig. 6.8 Javascript-based dynamic data table for exploring raw data
comments associated with the analyzed recommendation
option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 67
Fig. 7.1 Recommendations generated for client 32, category service,
2016 based on extracted rules of “Type 0” . . . . . . . . . . . . . . . .. 83
Fig. 7.2 Recommendations generated for client 32, category service,
2016 based on extracted rules of “Type 1” . . . . . . . . . . . . . . . .. 84
Fig. 8.1 Window interface for Recommender system presenting terms
used in the initial approach . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 89
Fig. 9.1 The illustration of the data collected within the currently
existing customer retention program to determine the customer
activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Fig. 9.2 Initial analysis of the number of surveys per year . . . . . . . . . . . 118
Fig. 9.3 Analysis of the number of distinct customers surveyed
per year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
List of Figures xv
Fig. 9.4 Analysis of the decision attribute—Customer

Status—distribution of different categories . . . . . . . . . . . . . . . . . 119
Fig. 9.5 Attribute selection using best first method . . . . . . . . . . . . . . . . . 120
Fig. 9.6 Attribute selection using information gain method . . . . . . . . . . . 121
Fig. 9.7 The setup of “Data Mining” task—action rule mining
with LISp-Miner to find the reasons behind the customer
defection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Fig. 9.8 The results of action rule mining in LISp-Miner. . . . . . . . . . . . . 122
List of Tables
Table 5.1 Sample meta-actions influence matrix . . . . . . . . . . . . . . . . . .. 55

Table 7.1 A summary of experiments on a small dataset (Client
24 - 306 surveys, rules Detractor to Promoter) . . . . . . . . . . . .. 75
Table 7.2 A summary of experiments on a small dataset (Client 5 - 213
surveys, rules Passive to Promoter) . . . . . . . . . . . . . . . . . . . .. 76
Table 7.3 A summary of experiments on a small dataset (Client 5 - 306
surveys, rules Passive to Promoter) . . . . . . . . . . . . . . . . . . . .. 76
Table 7.4 A summary of experiments on a medium dataset (Client
16 - 1192 surveys, rules Detractor to Promoter) . . . . . . . . . . .. 77
Table 7.5 A summary of experiments on a medium dataset (Client
16 - 1192 surveys, rules Passive to Promoter) . . . . . . . . . . . .. 77
Table 7.6 Comparison of running times in Spark and Java for datasets
from category parts 2015 . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 78
Table 7.7 Comparison of coverage in Spark and Java for datasets
from category parts 2015 . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 78
Table 7.8 Comparison of meta action mining, triggeration and meta
node creation processes for a small dataset with different
rule types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 81
node creation processes for a medium dataset with different
rule types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 82
node creation processes for a large dataset with different
rule types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 82
Table 8.1 Example of decision table built from expected opinion
mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 90
Table 8.2 Comparing results of different approaches to sentiment
analysis—metrics calculated per comment . . . . . . . . . . . . . . .. 98
Table 8.3 Comparing results of different approaches to sentiment
analysis—metrics calculated per opinion . . . . . . . . . . . . . . . .. 99
xvii
xviii List of Tables
Table 8.4 Improving sentiment analysis further for the combined

approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Table 8.5 Sparsity of opinion table before and after modifications
of the opinion mining algorithm—data for Client 16 . . . . . . . . 103
Table 8.6 Sparsity of opinion table before and after modifications
of the opinion mining algorithm—data for Client 3 . . . . . . . . . 103
Table 8.7 Results of the action rule mining on the opinion table
for Client 3 and Client 16 before and after modifications
to the opinion mining algorithm . . . . . . . . . . . . . . . . . . . . . . . . 104
Table 8.8 Comparison of recommendations results in the previous
and the new (text-only) approach for Client 16
and Client 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Chapter 1
Introduction
1.1 Why Customer Experience Matters More Now?
Customer experience! These are two words mentioned frequently in business con-
versations as well as the business press. To illustrate just how frequently, searching
customer experience in Google yields 1.3 billion results! By comparison, a search
for CRM results in only 157 million results; employee engagement 164 million; and
ERP (enterprise resource planning software) 168 million. Why the strong interest in
customer experience? Simply, a need to find more and better ways to differentiate
the products and services offered by a company. Consider some history. In the 1990s,
industrial managers spent a lot of time on product innovation. Having a new prod-
uct with innovative features was a proven way to win customers. The construction
equipment industry provides a great example. Prior to the early 1970s, bulldozers
had manual transmissions. In the early 1970s, JCB first introduced a bulldozer with a
hydrostatic transmission. This type of transmission dramatically improved machine
and operator productivity among other things. No longer was it necessary to stop,
engage a clutch and shift gears to change direction or change the speed. Now, these
changes could be done with a joy stick. Companies rapidly developed such offer-
ings and now, nearly all bulldozers sold now have either a hydrostatic or automatic
transmission. While there are differences in the transmissions offered by the various
manufacturers, providing a product that did not require manual shifting became a
required feature. In addition to illustrating the importance of product improvement
as a key business strategy, it also shows the limits of such a strategy. In a world where
most every supplier offers a similar set of product features, innovation becomes a
less potent tool for differentiation. Customer experience was seldom in many strate-
gic planning sessions discussed formally. Unless the customer experience was just
awful, managers assumed it was adequate. In short, poor service could, in some
cases, hinder a deal but great customer service was not seen as a tool to create more
loyal customers and cause them to talk with other potential customers. To illustrate
© Springer Nature Switzerland AG 2020 1

K. Tarnowska et al., Recommender System for Improving Customer Loyalty,
Studies in Big Data 55, https://doi.org/10.1007/978-3-030-13438-9_1
2 1 Introduction
the thinking of the time, consider the work of one of the leading business strategy
thinkers, Michael Porter. In 1979, he wrote an article entitled “The Five Competitive
Forces that Shape Strategy” (Harvard Business Review, January 2008). In the article,
he outlined five forces that help to shape business strategy. To compete against these
forces he suggested three broad strategies: cost leadership; differentiation; and focus.
When he discussed the differentiation strategy, it was largely about product differen-
tiation and/or cost superiority. The above is not a criticism of the profound impact that
Michael Porter’s strategic thinking has had on many business leaders. Rather, it illus-
trates the lack of attention paid to customer service as a strategic differentiator. In the
mid-2000s, this calculus changed, for two big reasons. First, design and engineering
capability and capacity increased throughout the world. For example, the number of
science and engineering degrees awarded in China increased from 359,000 in 2000
to 1.65 million in 2014 (article from Science Policy News of the American Insti-
tute of Physics, January 2018. The article summarizes the National Science Board’s
biennial Science and Engineering Indicators research). India has also been producing
large numbers of engineers and scientists as has Europe, Japan and the US. Consider
the current race between China and the US to build the 5G wireless network of the
future. Huawei Technologies, a Chinese company, had one of its components named
as a critical component to the 5G system and the developer of that technology is a
Turkish-born scientist named Erdal Arikan (Wall Street Journal, The 5G Race: China
and U.S. Battle to Control World’s Fastest Wireless Internet, September 9, 2018). 5G
Race: This huge increase in intellectual capital meant that product innovation was
no longer primarily found in the US, Japan or Europe. New and innovative products
began appearing from a wide variety of countries. Couple a first-rate product design
with competitive production costs and you have a strong value proposition. Consider
what Hyundai and Kia (now part of the same company), South Korean automotive
companies have done. While some may argue that the products designs are uninspir-
ing, both offer a line of automobiles that are well-engineered and attractively priced.
US sales are growing. Second, during this period, the internet came to the fore. The
dotcom bubble in the late 1990s resulted in a massive overinvestment in high-speed
fiber optic cable. This high-speed capacity made it easier to get information about
products and companies anywhere in the world. Thomas Friedman wrote about it
most eloquently in the World is Flat, when he said it leveled the playing field. While
product differentiation was (and is) an important strategy for any company, it has
become one of several important strategies and not the primary one. With the advent
of more and stronger technical skills in many parts of the world, a differentiation
strategy depending almost exclusively on product innovation is less defensible than
in the past. With the plethora of information of all types, potential buyers can find out
if a product’s technical superiority is really all that superior! Critically, they can find
out what existing users think about a product and the service it receives. Changing
customer expectations have also played a critical role in making customer experience
more important in the business-to-business arena. The business-to-consumer arena
played a role in this as did the increasing complexity of many of the products and
1.1 Why Customer Experience Matters More Now? 3
services sold in the business-to-business arena. How did the B-to-C market influence
B-to-B? One big example is Amazon. The company made it far easier to find prod-
ucts, get user reviews of those products and then order them. Consumers liked it (as do
shareholders since Amazon now has a market cap of over one trillion dollars). There
are many other examples in addition to Amazon but suffice it to say, that they made
it quicker and easier to get consumer products? Those same B-to-B buyers were also
consumers of a range of products they purchased online. They realized how easy and
frictionless it was, which influenced their expectations when it came to purchasing
products and services for the businesses in which they worked. To illustrate, visit
most any industrial products company website and they will have a link to a parts
store or, in come cases, the option to order online. Rather than picking up the phone
or meeting with a salesperson, customers want the ease of ordering online, where
possible, much in the same way they do when they order a product from Amazon.
Increasing product complexity has also created an increased need for improving the
customer experience. Electronic technology of all types is increasingly integrated
into products. While the benefits of this technology are quite significant there are
increased product support needs. The typical higher horsepower farm tractor now
comes with a climate-controlled cab that contains two or three electronic displays.
These displays are monitoring tractor performance (e.g., engine, transmission, etc.),
attachment functions (e.g., planter placing seeds accurately), and another that shows
the GPS location so that the operator can turn the tractor at the end of the row, get it in
the correct position and let the GPS system drive it to the end of the row (the operator
is only monitoring). Compare this with a tractor of 20 years ago that may have had
a simple cover for operator protection from the elements and no monitors. A very
different picture than what is happening today. This technology is a real productiv-
ity booster but learning how to use it and keeping it maintained require significant
after-sale support. When a customer spends hundreds of thousands of dollars on a
piece of equipment, keeping it running and running efficiently is critical to realizing
the desired return on investment.
1.2 Top (and Bottom) Line Reasons for Better Customer

Experience
The previous few paragraphs outline key macro reason for why a product differ-
entiation strategy became less strategically potent. At the micro level, managers
discovered that having better customer experience pays in many ways. For example:
1. A research project showed that the higher the satisfaction the greater increase
in sales. Customers that had the best past experiences spent a 140% more than
customers with lower satisfaction ratings (The Value of Customer Experience
Quantified, Harvard Business Review, August 1, 2014).
4 1 Introduction
2. Internal research conducted by The Daniel Group shows that more satisfied
customers actively refer. In the farm equipment market, they found that about
40% of farmers indicated they gave a referral for a dealer in the past 6 months.
However, over 90% of the referrals came from the most satisfied customers.
3. A study of publicly traded companies by Watermark Consulting showed that
from 2007 to 2013, those companies with better customer service generated a
total return to shareholders that was 26 points higher than the S and P 500.
Managers today are learning that creating meaningful differentiation is about offer-
ing value in different ways and not just one or two. AGCO, a global manufacturer
of agricultural equipment just recently introduced its new row-crop tractor
(http://www.challenger-ag.us/products/tractors/1000-series-high-horsepower-row-
crop-tractors.html). The tractor offers an innovative design with many benefits to
its users (e.g., buy one tractor that can do the job of two) and it is proving to be a
very high-quality product. Simultaneously, AGCO is working to improve the product
support its dealers are providing buyers. Based on initial results, this multi-pronged
strategy is working. The product is getting good reviews. Any issues that customers
are having in the field are being promptly handled by AGCO and its dealer network.
An improved customer experience matters for many financial reasons, as noted pre-
viously. If one of a company’s strategies is built around product differentiation, then,
as the AGCO illustration shows, a better service experience enhances the chances of
success for product differentiation. What problems does the recommender system
address? The recommender system addresses several important issues. The most
important are:
1. Provides a decision framework so managers can understand which action or set
of actions are likely to have the greatest impact on the Net Promoter Score (NPS).
Managers often know they need to improve customer experience but are often
lost as to where to begin. If managers need to improve inventory turns, as an
example of another exercise they may do, there are pieces of information that can
help guide and inform decision-making (e.g., identify inventory items with low
turns and high investment). This information helps them to identify priorities for
action. When faced with improving customer experience, there are few similar
pieces of information to inform the decision. For example, if a difficult-to-use
phone system is raised by customers, what happens to likelihood to recommend
if the phone system is improved? The recommender system can provide insights
as to what the likely NPS improvement is likely to be if the phone systems are
improved. This improves managerial decision-making.
2. Results are based on multiple clients. The data mining techniques deployed in
the recommender system allow for learning to be gained based on the experi-
ence of other users, without sharing proprietary information. This dramatically
strengthens the power of the system.
1.2 Top (and Bottom) Line Reasons for Better Customer Experience 5
3. Strengthens traditional text mining options. Text mining can be useful to identify
the frequency with which topics are mentioned and the sentiment associated with
the topics. The recommender system allows users to see specific anonymous com-
ments associated with actual customers. Studying these comments can provide
very granular insights into steps that can be taken to improve customer experi-
ence. In addition to quantifying the potential impact on NPS of various actions,
it allows users to better understand the specific things that when done, improve
NPS and, therefore, the overall customer experience.
4. Provides a sensitivity analysis. In some cases, certain actions or sets of actions are
more easily implemented than others. The recommender system allows managers
to weight these actions to determine which ones have more impact. For example,
one action may be to conduct training to improve communication. While it may be
shown to positively impact NPS, it may not be very practical in large and sprawling
organizations. The system allows the various sets of actions to be weighted for
feasibility, which enhances the managerial decision-making process.
1.3 What is Next?
We worked on this system for a few years and it was a great learning experience for
all of us. But there is more to be done. Our work showed that it is possible to create a
rigorous framework by which to analyze various actions to improve NPS. This helps
improve decision-making for managers tasked with improving service experience.
This framework needs to be further enhanced and strengthened to identify even
better ways to derive meaning from textual comments and more deeply understand
the impact on NPS of various actions. We also need to learn how this approach applies
to other markets and industries. Our suspicion is that it is very applicable. The firm
we collaborate with works in a variety of industrial markets and we often see similar
issues no matter the industry. For example, poor communication is one of those
ubiquitous problems that frequently shows up as something that negatively impacts
customer experience in a variety of industries. It would be interesting and useful
to see how customers in different industries respond to similar actions to improve
communication.
1.4 Final Observations
Customer experience is a very robust element in the strategic arsenal of a company.

While product innovation is important, it is also becoming more challenging to
implement due to greater information flow and a much larger pool of creative talent
throughout the world. However, a company that provides easy, reliable customer
service is a tough competitor. A company with great customer experience has a very
defensible strategy that is difficult to challenge. There is still much to be learned about
6 1 Introduction
effective and powerful ways to improve customer experience. The recommender

system provides some powerful insights into how this can be done.
Chapter 2
Customer Loyalty Improvement
2.1 Introduction to the Problem Area
Nowadays most businesses, whether small-, medium-sized or enterprise-level orga-

nizations with hundreds or thousands of locations collect their customers feedback
on products or services. A popular industry standard for measuring customer satis-
faction is called “Net Promoter Score”1 [1] based on the percentage of customers
classified as “detractors”, “passives” and “promoters” (see Fig. 2.1). Promoters are
loyal enthusiasts who are buying from a company and urge their friends to do so.
Passives are satisfied but unenthusiastic customers who can be easily taken by com-
petitors, while detractors are the least loyal customers who may urge their friends
to avoid that company. The total Net Promoter Score is computed as %Promoters
-%Detractors. The goal here is to maximize NPS, which in practice, as it turns out, is
a difficult task to achieve, especially when the company has already quite high NPS.
Most executives would like to know not only the changes of that score, but also
why the score moved up or down. More helpful and insightful would be to look
beyond the surface level and dive into the entire anatomy of feedback.
The main problem to solve is to understand difference in data patterns of customer
sentiment on a single client personalization level, in years 2011–2016. The same it
should enabled to explain changes, as well as predict sentiment changes in the future.
Actionable knowledge is needed for the business for designing the strategic directions
that would help drive customer loyalty improvement.
1 NPS®, Net Promoter®and Net Promoter®Score are registered trademarks of Satmetrix Systems,
Inc., Bain and Company and Fred Reichheld.
8 2 Customer Loyalty Improvement
Fig. 2.1 The concept of Net

Promoter Score as a way to
quantify and categorize
customer satisfaction
2.2 Dataset Description
The chosen dataset is related to a research project conducted in the KDD Lab at
UNC-Charlotte in collaboration with a consulting company based in Charlotte. The
company collects data from telephone surveys on customer’s satisfaction from re-
pair service done by heavy equipment repair companies (called clients). There are
different types of surveys, depending on which area of customer satisfaction they
focus on: Service, Parts, Rentals, etc. The consulting company provides advisory for
improving their clients’ Net Promoter Score rating and growth performance in gen-
eral. Advised clients are scattered among all the states in the US (as well as Canada)
and can have many subsidiaries. There are above 400,000 records in the dataset in
total (years 2011–2016), and the data is kept being collected. The dataset (Fig. 2.2)
consists of features related to:
1. Clients’ details (repair company’s name, division, etc.);
2. Type of service done, repair costs, location and time;
3. Customer’s details (name, contact, address);
4. Survey details (timestamp, localization) and customers’ answers to the questions
in survey;
5. Each answer is scored with 1–10 (optionally textual comment) and based on total
average score (PromoterScore) a customer is labeled as either Promoter, Passive
or Detractor of the given client.
The data is high-dimensional with many features related to particular assessment
(survey questions) areas, their scores and textual comments. The consulted clients as
Fig. 2.2 Illustration of the NPS dataset structure—the features and the decision attribute
2.2 Dataset Description 9
well as surveyed customers are spread geographically across United States. Records
are described with temporal features, such as DateInterviewed, InvoiceDate and
WorkOrderCloseDate.
2.3 Decision Problem
The first goal is to find characteristics (features) which most strongly correlate with
Promoter/Detractor label (PromoterScore), so that we can identify areas, where
improvement can lead to changing a customer’s status from Detractor to Promoter
(improvement of customer’s satisfaction and client’s performance). Identifying these
variables (areas) helps in removing redundancy.
So far correlation analysis has considered only global statistics, however global
statistics can hide potentially important differentiating local variation. The problem
is multidimensional as the correlations vary across space, with scale and over time.
The intermediate goal is also to explore the geography of the issue and use inter-
active visualization to identify interdependencies in multivariate dataset. It should
support geographically informed multidimensional analysis and discover local pat-
terns in customers’ experience and service assessment. Finally, classification on NPS
should be performed on semantically similar customers (similarity can be also based
on geography). A subset of most relevant features should be chosen to build a clas-
sification model.
2.4 Problem Area
The following subsections present the most important problem areas identified within
the research.
2.4.1 Attribute Analysis
The first problem that needs to be solved is to find out which benchmarks are the most
relevant for Promoter Status. There is also a need to analyze how the importance of
benchmarks changed over years for different clients (locally) and in general (glob-
ally), and additionally how these changes affected changes in Net Promoter Score,
especially if this score deteriorated (which means customer satisfaction worsened).
There is a need to identify what triggered the highest NPS decreases and the highest
NPS growths.
The business questions to answer here are:
• What (which aspects) triggered changes in our Net Promoter Score?
10 2 Customer Loyalty Improvement
• Where did we go wrong? What could be improved?

• What are the trends in customer sentiment towards our services? Did more of
them become Promoters? Passives? Detractors? Did Promoters become Passives?
Promoters become Detractors? If yes, why?
The problem with the data is that the set of benchmarks asked is not consistent and
varies for customers, clients and years. Customer expectations change as well. There-
fore, one has to deal with a highly incomplete and multidimensional data problem.
2.4.2 Attribute Reduction
The consulting company developed over 200 such benchmarks in total (223 in 2016),
but taking into considerations time constraints for conducting a survey on one cus-
tomer it is impossible to ask all of them. Usually only some small subsets of them
are asked in a survey. There is a need for a benchmark (survey) reduction, but it is
not obvious which one of them should be asked so that to obtain the most insightful
knowledge. For example, consulting aims to reduce the number of questions to the
three most important, such as “Overall Satisfaction”, “Referral Behavior” and “Pro-
moter Score”, but it has to be checked if this will not lead to significant knowledge
loss about customer satisfaction problem. There is a need to know which bench-
marks can/cannot be dropped in order to decrease a knowledge loss. For example, in
years 2014–2015 some questions were asked less frequently because questionnaire
structure changed and survey shortened for some clients. There is a need for analysis
regarding how these changes in the dataset affect the previously built classification
and recommender system model.
2.4.3 Customer Satisfaction Analysis and Recognition
The second application area is tracking the quality of the knowledge base being
collected year by year, especially in terms of its ability to discern between differ-
ent types of customers defined as Promoters, Passives and Detractors. The main
questions business would like to know the answers to are:
• What (which aspect of service provided) makes their customers being Promoters
or Detractors?
• Which area of the service needs improvement so that we can maximize customer
satisfaction?
For every client company the minimum set of features (benchmarks) needed to clas-
sify correctly if a customer is a Promoter, Passive or Detractor should be identified.
The strength of these features and how important a feature is in recognition process
needs to be determined. However, answering these questions is not an easy task, as
2.4 Problem Area 11
the problem is multidimensional, varies in space and time and is highly dependent on
the data structure. Sufficient number of customer feedback on various aspects must
be collected and analyzed in order to answer these questions. Often human abilities
are not sufficient to analyze such huge volume of data in terms of so many aspects.
There is a need for some kind of automation of the task or visual analytics support.
2.4.4 Providing Recommendations
The main goal of this research work is to support consulting business with recom-
mendations (recommendable sets of actions) to their clients (repair companies), so
that they can improve their NPS. The items must be evaluated in terms of some
objective metrics.
Besides, the recommendation process needs to be more transparent, valid and
trustworthy. Therefore, the need to visualize an algorithm process which leads to
generating a recommendation output. The end user must be able to understand how
recommendation model works in order to be able to explain and defend the model
validity. Visual techniques should facilitate this process.
Reference
1. SATMETRIX. Improving your net promoter scores through strategic account management.
http://info.satmetrix.com/white-paper-download-page-improving-your-net-promoter-scores-
through-strategic-account-management. Accessed: 2017-04-26.
Chapter 3
State of the Art
In this chapter, different types of available information technology solutions support-

ing customer relationship management as well as collecting the customer feedback,
are discussed, with the focus on the new generation on intelligent decision support
and recommender systems.
3.1 Customer Satisfaction Software Tools
Horst Schulz, former president of the Ritz-Carlton Hotel Company, was famously
quoted as saying: “Unless you have 100% customer satisfaction…you must im-
prove”. Customer satisfaction software helps to measure customers’ satisfaction as
well as gain insight into ways to achieve higher satisfaction. SurveyMonkey [1] is the
industry leading online survey tool used by millions of businesses across the world.
It helps to create any type of survey, but it also lacks features with regard to measur-
ing satisfaction and getting actionable feedback. Client Heartbeat [2] is another tool
built specifically to measure customer satisfaction, track changes in satisfaction lev-
els and identify customers ‘at risk’. SurveyGizmo [3] is another professional tool for
gathering customer feedback. It offers customizable customer satisfaction surveys,
but it also lacks features that would help to intelligently analyze the data. Customer
Sure [4] is a tool that focuses on customer feedback: facilitates distribution of cus-
tomer surveys, gathering the results. It allows to act intelligently on the feedback
by tracing customer satisfaction scores over time and observe trends. Floqapp [5] is
a tool that offers customer satisfaction survey templates, collects the data and puts
it into reports. Temper [6] is better at gauging satisfaction as opposed to just being
a survey tool. Similar to Client Heartbeat, Temper measures and tracks customer
satisfaction over a period of time.
These types of tools mostly facilitate design of surveys, however, offer very lim-
ited analytics and insight into customer feedback. It mostly confines to simple trend
analysis (tracing if score increased or decreased over time). The Qualtrics Insight
Platform [7] is a leading platform for actionable customer, market and employee
14 3 State of the Art
insights. Besides customers’ feedback collection, analysis and sharing it offers ex-
tensive insight capabilities, including tools for end-to-end customer experience man-
agement programs, customer and market research and employee engagement.
3.2 Customer Relationship Management Systems
CRM is described as “managerial efforts to manage business interactions with cus-

tomers by combining business processes and technologies that seek to understand
a company’s customers” [8], i.e. structuring and managing the relationship with
customers. CRM covers all the processes related to customer acquisition, customer
cultivation, and customer retention. CRM also involves development of the offer:
which products to sell to which customers and through which channel. CRM seeks
to retain customers and design marketing campaigns. Sometimes CRM strategy is
incorporated into other enterprise systems. An enterprise data warehouse has become
a critical component of a successful CRM strategy [9]. Data mining techniques in this
area are useful for extracting marketing knowledge and further supporting marketing
decisions. The CRM systems must analyze the data using statistical tools and data
mining. There are two critical components of marketing intelligence: customer data
transformation and customer knowledge discovery.
3.3 Decision Support Systems
A DSS is an interactive computer-based system designed to help in decision making

situations by utilizing data and models to solve unstructured problems [8].
The aim of DSSs is to improve and expedite the processes by which manage-
ment makes and communicates decisions—in most cases the emphasis in DSSs is
on increasing individual and organizational effectiveness. DSS in general can im-
prove strategic planning and strategic control. Research indicates data-driven or data-
informed organizations improve decision-making, increase profitability and drive
innovation. As strategic planning requires large amount of information, the only
effective way to manage large amounts of information is with information technolo-
gies. Proper integration of DSSs and CRM presents new opportunities for enhancing
the quality of support provided by each system.
3.4 Recommender Systems
Most recommender systems were applied in e-commerce settings, supporting cus-

tomers in online purchases of commodity products such as books, CDs. Idea of
applying recommender system in the area of strategic business planning seems to
Another random document with
no related content on Scribd:
being decidedly in a minority. This alone would be sufficient to
stamp the festival as one belonging peculiarly to the women.
The structure where this ceremony was to take place was typically
African, not over large, but quite sufficiently so for the object in view.
The natives thoroughly understand the art of putting up buildings
admirably suited to the purpose they are to serve, and also quite
pleasing in style and shape, out of the cheapest materials and with
the simplest appliances, in a very short time. This hut was circular,
with an encircling wall of poles and millet-straw, between six and
seven feet high. It was about thirty feet in diameter, with two
doorways facing each other, and a central post supporting the roof.
The women were just entering in solemn procession, while the
tuning up of several drums was heard from the inside. The jumbe’s
hint as to the shyness of the women was abundantly justified; those
who caught sight of us at once ran away. The participants only grew
calm when we had succeeded in getting up unseen close to the outer
wall of the building and there finding shelter in a group of men
disposed to be sensible. It was, however, even now impossible to
sketch any of the women. I am in the habit, wherever I can, of jotting
down in a few rapid strokes every picturesque “bit” I come across,
and here I found them in unusual number. Since I left the coast,
labrets, nose-pins, and ear-studs have become quite hackneyed, but
hitherto I had come across no specimens of such size or racial types
so markedly savage and intact. When one of these women laughs, the
effect is simply indescribable. So long as her face keeps its normal
serious expression, the snow-white disc remains in a horizontal
position, that is to say, if the wearer is still young and good-looking.
If, however, she breaks into the short, giggling laugh peculiar to the
young negress, the pelele flies up with an abrupt jerk and stands
straight up over the ivory-white and still perfect teeth, while the
young woman’s pretty brown eyes flash with merriment, and the
weight of the heavy wooden plug sets up a quick vibration in the
upper lip, which is dragged out by almost a hand-breadth from its
normal position. Then the baby on the woman’s back (nearly all of
them are carrying babies), begins to cry piteously under the
searching gaze of the strange white man; and, in short, the whole
spectacle is one which must be seen to be appreciated—no pen can
describe it.
Our place was well chosen, and enabled us to survey the whole
interior of the hut without let or hindrance. I noticed three youths
sitting on stools of honour in a reserved part of the hall, and inquired
of the jumbe, who stood beside me, obligingly ready to be of use, who
those three little shrimps were? It appeared that they were the
husbands of the girls whose chiputu was being celebrated that day.
LAUGHING BEAUTIES
And what is chiputu? It is the celebration of a girl’s arrival at

womanhood; but that is a long story, which we have no time to
investigate just now, for the drums have struck up, in that peculiar
cadence, heard at every ngoma, which no one who has visited East
Africa can ever forget. At the same moment the closely-packed
throng of black bodies has already arranged itself for a dance. With a
step something like the gait of a water-wagtail, they move,
rhythmically gliding and rocking, round the central posts, at which
three old hags stand grinning.
“Who are those?” I ask.
Those are the anamungwi, the instructresses of the three girls;
they are to receive the reward of their work to-day. “See now, sir,
what is happening.” For the moment nothing happens, the dance
goes on and on, first in the way already described, then changing to
one which is not so much African as generally Oriental: it is the so-
called danse du ventre. At last this too comes to an end, the figure
breaks up in wild confusion, one snatching in this direction, another
in that, and everyone gathers once more round the anamungwi.
These are no longer smiling, but comport themselves with great
dignity as they have every right to do. One after another, the women
come forward to hand them their gifts, pieces of new cloth, strings of
beads, bead necklaces and armlets, and various items of a similar
character. “That is all very fine,” their looks seem to say, “but is this
an equivalent for the unspeakable trouble which the training of our
amwali, our pupils, has given us for years past? We expect
something more than that!” However, the festive throng are not in
the least disturbed by this mute criticism; people all chatter at once,
just as they do in other parts of the world, and everyone is in the
highest spirits.
GIRLS’ UNYAGO AT THE MAKONDE HAMLET OF NIUCHI
Now comes a new stage. “Hawara marre” mutters the jumbe. This
even Nils Knudsen cannot translate, for it is Kimakua, which he does
not know, but the jumbe, like all intelligent men in this country, is a
polyglottist. He says the Yao for it is “Chisuwi mkamule” (“The
leopard breaks out”). At this moment something unexpected
happens. The three young fellows rise quick as lightning, and, with
loud crashing and rustling, they have burst through the fragile hut-
wall and are seen retiring towards the outskirts of the village. I have
not yet clearly made out whether these youthful husbands
themselves represent the leopard or whether they are to be thought
of as pursued by an imaginary leopard. In either case, the leisurely
pace at which they stroll away is scarcely convincing and still less
imposing; less so, certainly, than the song of Hawara marre,
rendered by the women with equal spirit and energy, which rings out
into the sun-baked pori long after the three leopards have vanished
in the distance.
Now comes another picture; the hall is empty, but the open space
beside it, which has been carefully swept, swarms with brightly-
coloured fantastic figures. It is only now that we can see how they
have adorned themselves for the occasion. The massive brass
bangles, nearly an inch thick, which they wear on their wrists and
ankles, shine like burnished gold, and the calico of their skirts and
upper garments is of the brightest colours. These cloths, in fact, have
just been bought from the Indian traders at Lindi or Mrweka, at
great expense, by the gallant husbands, who have recently made an
expedition to the coast for the purpose. The white pelele seems to
shine whiter than usual, and the woolly heads and brown faces are
quite lustrous with freshly-applied castor oil, the universal cosmetic
of these regions. Once more the anamungwi take up a majestic pose,
and once more all the women crowd round them. This time the
presents consist of cobs of maize, heads of millet, and other useful
household supplies, which are showered wholesale on the recipients.
Once more the scene changes. The drummers have been tuning up
their instruments more carefully than usual, and at this moment the
fire blazes up for the last time and then expires. The first drum
begins—boom, boom, boóm, boom, boom, boóm, boom, boom,
boóm: two short notes followed by a long one. How the man’s hands
fly! There are more ways of drumming than one, certainly,—but the
art as practised here seems to require a special gift. It is by no means
a matter of indifference whether the drumhead is struck with the
whole hand, or with the finger-tips only, or whether the sound is
produced by the knuckles or finger-joints of the closed fist. It is
pretty generally assumed that we Europeans have an entirely
different mental organization from that of the black race, but even
we are not unaffected by the rhythm of this particular kind of
drumming. On the contrary, the European involuntarily begins to
move his legs and bend his knees in time to the music, and would
almost feel impelled to join the ranks of the dancers, were it not for
the necessity of maintaining the decorum of the ruling race, and of
keeping eye and ear on the alert for everything that is going forward.
The dance which the women are now performing is called ikoma.
[41]
Our eyes are insufficiently trained to perceive the slight
differences between these various choric dances, and so we grew
tired with mere looking on long before the natives, who are exerting
themselves to the utmost, begin to weary. In this case the sun
contributes to the result, and Moritz is already feeling ill, as he says,
from the smell of the crowd; though he certainly has no right to look
down on his compatriots in this respect. It is true that he has
improved since the day at Lindi, when I drove him before my kiboko
into the Indian Ocean, because he diffused around him such a
frightful effluvium of “high” shark, that it seemed as if he himself had
been buried for months. I am just about to pack up my apparatus,
when the uniform, somewhat tedious rhythm in which the crowd of
black bodies is moving suddenly changes. Hitherto, everything has
been characterized by the utmost decency, even according to our
standards, but now what do I see? With swift gesture the bright-
coloured draperies fly up, leaving legs and hips entirely free, the feet
move faster, and with a more vivacious and rapid motion the dancers
now circle round one another in pairs. I am fixed to the spot by a
sight I have often heard of, but which has never come in my way
before:—the large keloids which, in the most varied patterns cover
these parts of the body. The scars are raised to this size by cutting
again and again during the process of healing. This, too, belongs to
the ideal of beauty in this country.
Unfortunately, I was not able to await the end of the ikoma. The
performers, in spite of the small silver coin which I had distributed to
each of them, were evidently constrained in the presence of a
European,—a being known to most of them only by hearsay—and the
spontaneous merriment which had prevailed inside the hut was not
to be recovered. Besides, I was forced, out of consideration for
Moritz, who was now quite grey in the face, to return as quickly as
possible.
Akundonde’s junior headman is excellent as a practical guide, but
has little theoretic knowledge,—he is probably too young to know
much of the traditional lore of his own tribe and the Makua. Old
Akundonde himself keeps silence,—perhaps because he needs a
stronger inducement than any yet received. This, however, I am
unable to offer, especially as we ourselves have to subsist on our
tinned goods, the usual lean fowls and a few old guinea-fowl shot by
Knudsen. There is no trace of the liberal gifts of pombe which had
delighted our thirsty souls at Masasi and Chingulungulu.
It was, therefore, with light hearts that we left Akundonde’s on the
fourth day for Newala. The stages of our three days’ march were
Chingulungulu, where we had left a considerable part of our baggage,
and Mchauru, a very scattered village in a district and on a river of
the same name, in the foothills of the Makonde plateau. Mchauru is
interesting enough in several respects. First, topographically: the
river, which has excavated for itself a channel sixty, in some places
even ninety feet deep, in the loose alluvial soil, runs south-westward
towards the Rovuma. On reaching the bottom of this gorge, after a
difficult climb, we found no running water, but had to dig at least a
fathom into the clean sand before coming on the subterranean
supply. The deep, narrow water-holes, frequently met with show that
the natives are well aware of this circumstance. The vegetation in this
whole district, however, is very rich, and it is not easy to see at
present whence it comes, since we are on the landward side of the
hills whose seaward slope precipitates the rains. It is possible that
the soil here holds more moisture than in other parts of the plain.
Mchauru has not only charming scenery but abounds in
ethnographic interest. It possesses, in the first place, a fundi who
makes the finest ebony nose-pins in the country, and inlays them
with zinc in the most tasteful manner, and secondly, a celebrated
magician by the name of Medula. In fact, it was on account of these
two men that I halted here at all. The nose-pin-maker was not to be
found—we were told that he was away on a journey—but Medula was
at home.
From our camp, pitched under a huge tree beside the road, we—
that is Knudsen and I, with my more immediate followers carrying
the apparatus—walked through banana groves (which I now saw for
the first time), and extensive fields of maize, beans, and peas, ready
for gathering, in a south-westerly direction for nearly an hour. At
intervals the path runs along the bed of a stream, where the deep
sand makes walking difficult. At last, on ascending a small hill, we
found ourselves before an open shed in which an old native was
seated, not squatting in the usual way, but with his legs stretched out
before him, like a European. After salutations, my errand was
explained to him,—I wanted him to tell me all about his medicines
and sell me some of them, also to weave something for us. According
to native report, there are only two men left in the whole country
who still possess this art, already obsolete through the cheapness of
imported calico. Medula is one of these weavers,—the other, a
tottering old man, I saw, several weeks ago, at Mkululu. I was greatly
disappointed in him; he had not the faintest notion of weaving, and
there was nothing in the shape of a loom to be seen in his hut; the
only thing he could do was to spin a moderately good cotton thread
on the distaff.
PARTICIPANTS ASSEMBLING AT THE UNYAGO HUT
PRESENTATION OF CALICO BY THE MOTHERS

DANCE OF THE OLD WOMEN
ARRIVAL OF THE NOVICES
GIRLS’ UNYAGO AT THE MATAMBWE VILLAGE OF MANGUPA. I
OLD WOMEN GROUPED ROUND THE GIRLS TO BE

INITIATED
DANCE OF THE OLD WOMEN ROUND THE INITIATES
DANCE OF THE INITIATES BEFORE THE OLD WOMEN
DEPARTURE OF THE INITIATES
GIRLS’ UNYAGO AT THE MATAMBWE VILLAGE OF MANGUPA.

II
OLD MEDULA LIGHTING HIS PIPE
I expected more satisfactory results from Medula; but the

medicines were the first point to be attended to. We haggled with
him like Armenians, but he would concede nothing, finally showing
us one or two of the usual calabashes with their questionable
contents, but demanding so exorbitant a price that it was my turn to
say, as I had great satisfaction in doing, “Hapana rafiki” (“It won’t
do, my friend”). Medula is a philosopher in his way—“Well, if it
won’t, it won’t,” appeared to be his reflection, as he turned the
conversation to the subject of his name, then tried to pronounce
mine, and gradually passed over to the second part of our
programme. All this time I was on the watch with my camera, like
the reporter of some detestable illustrated weekly. Medula was
seated in an unfavourable position: bright light outside—deep
shadow within his cool hut. I requested him to change his seat—he
declined. My entreaties and flatteries had no other result than to
make him grin, deliberately get out his pipe, light it with a burning
coal, and puff away without moving. Trusting to my Voigtländer’s
lens, I at last let him alone, as things had come to a standstill, and I
wanted to see the loom and its use. Medula said that he must first
make the thread. I submitted; the old man put a leisurely hand into a
basket, deliberately took out a handful of cotton-seeds, husked them
secundum artem and began beating the flaky white mass with a little
stick. In a surprisingly short time a fairly large quantity of cotton was
reduced to the proper consistency; Medula seized it in his left hand
and began to pull out the thread with his right. So far the process
looked familiar; the people who came over every winter during my
boyhood from Eichsfeld to our Hanoverian village, to spin the
farmers’ wool for them, always began in the same way. The parallel,
however, ceased with the next step, and the procedure became
entirely prehistoric. The new thread was knotted on to the end of
that on the distaff, the latter drawn through a cleft which takes the
place of the eye on our spinning-wheel, the spindle whirled in the
right hand, the left being extended as far as possible—and then both
arms moved downward; the spindle was quickly rolled round on the
upper part of the thigh, and the thread was ready for winding.
Medula contrived to weary us out with this performance, but never
produced his loom, in whose existence I have entirely ceased to
believe. He promised at our parting—which was marked by a decided
coolness—to bring the implement with him to Newala; but not even
the most stupid of my men gave any credit to his assurance.
CHAPTER XII
UNYAGO EVERYWHERE
Newala, middle of September, 1906.
The charming festival recently witnessed at Achikomu’s seems to

have broken the spell which debarred me, just when the season was
at its height, from gaining an insight into this most important and
interesting subject. In the short period since my arrival at Newala, I
have been present at no less than two typical celebrations, both of
them girls’ unyagos. This I owe to the kindness of the Akida Sefu.
Sefu bin Mwanyi is an Arab—apparently of unmixed blood—from
Sudi. He is a tall, light-complexioned man, with finely-cut features.
He knows a number of languages, excelling even Knudsen in this
respect, and I cannot say enough of the obliging way in which he has
endeavoured to further my plans ever since my arrival.
After a fatiguing climb up the edge of the cliff bordering the
plateau, which just at Newala is particularly steep, and a short rest,
we made hasty arrangements for encamping in the baraza—open as
usual to the dreaded evening wind—within the boma or palisade of
stakes. The cold that night was almost Arctic, and we wrapped
ourselves in all the blankets we could find. In the early dawn, the
zealous akida came in a great hurry, to conduct us to the Makua
village of Niuchi, where the concluding ceremony of the girls’ unyago
was fixed for that day, and where I was sure to see and hear much
that was new. An hour later, our party, this time including my mule,
had already wound its way through a long stretch of primæval
Makonde bush. It proved impossible to ride, however—the path,
bordered by thick, thorny scrub, being never two feet wide in the
most frequented parts. We suddenly walked out of the thickest bush
on to a small open space surrounded by houses, and perceived with
some astonishment a large crowd of strange-looking female figures,
who were staring at us, struck dumb with terror. I saw at once that,
here, too, it would be well to keep as much as possible in the
background, and disappeared with my men and all the apparatus
behind the nearest hut. From this coign of vantage, I was able to
watch undisturbed a whole series of performances which few if any
travellers, probably, have seen in exactly the form they here
assumed.
OUR CAMP AT NEWALA
It is eight in the morning; the Makonde bush, which almost closes

over our heads, is clad in the freshest green, one large tree in the
middle of the bwalo[42] and a few others of equal proportions rise
above the general level of the pori, and the low Makonde huts stand
out sharply in the clear morning air. The few women whom on our
arrival we found sweeping the bwalo with bunches of green twigs,
have vanished like lightning in the crowd surrounding five other
figures dressed in gaudy cloths. These are squatting in the shadow of
a hut, covering their eyes and temples with their hands, and staring
fixedly at the ground through their fingers. Then a shrill sound is
heard, and five or six women are seen hurrying with grotesque jumps
across the open space. As they raise the traditional cry of rejoicing,[43]
the pelele, here of truly fabulous dimensions, stands up straight in
the air, while the tongue, stretched out under it, vibrates rapidly to
and fro in the manner indispensable to the correct production of the
sound. The first six are soon followed by a dozen other women,
among whom one voice sings:—“Anamanduta, anamanduta, mwan-
angu mwanagwe” (“They go away, they go away, my dear child,”)—
the rest repeating the line in chorus. The song is accompanied by
accurately-rhythmical hand-clapping, as the dancers move in short
tripping steps backward and forward. “Surely a barbaric lament over
a parting,” I reflect, on hearing Sefu’s rapid translation, but already a
new song is heard:—
“Namahihio achikuta kumaweru” (“The
owl cries in the gardens”). This, too, is
repeated for some time, then once more, all
crowd round the five bundles of cloth. Five
elderly women now step forward out of the
throng and decorate the heads of their pupils
—for such are the gaudily-attired beings—with
bunches of millet. The latter now rise, and
take up their position in Indian file, each with
her hands on the shoulders of the one before
her. The drums strike up—old and young
together swaying with skilled vibration in the
danse du ventre.
“Chihakatu cha Kuliwile nandu kuhuma
nchere.” (“The chihakatu (small flat basket) of
Liwile is carried out of the house early.”) This
THE AUTHOR IN
WINTER COSTUME AT
is the song now chanted as before by solo and
NEWALA chorus. By the chihakatu is probably meant
the decoration of millet-heads—the natives
are fond of symbolical expressions.
This song in its turn comes to an end; the ranks of the dancers
break up and the women hasten in all directions, coming back to lay
further supplies of millet, manioc, cloth, etc., at the feet of the five
instructresses. These, meanwhile, have been preparing for the next
step. An egg is broken, a little of the yolk is rubbed on the forehead of
each girl and the rest mixed with castor oil and used to anoint the
girls on chest and back. This is the sign that they have reached
maturity, and that the unyago is over. The first part of the festival is
concluded by the presentation of more new cloth to the girls.
Sefu now points out to me a stick planted in the ground, and tells
me that medicines belonging to the unyago have been buried under
it. He also says that some months ago, a large pot of water was
buried at another spot in the bwalo; this was also “medicine.”
While I am listening to this explanation, the women have once
more taken their places. With a ntungululu which, even at the
distance at which we are standing, is almost enough to break the
drums of our ears, all the arms fly up with a jerk, then down again,
and the performers begin to clap their hands with a perfection of
rhythm and uniformity of action seemingly peculiar to the dwellers
on the shores of the Indian Ocean, in order to accompany the
following song:—
“Kanole wahuma kwetu likundasi kuyadika kuyedya ingombe.”
The meaning is something like this: “Just look at that girl; she has
borrowed a bead girdle, and is now trying to wear it gracefully and
becomingly.”
Women are very much alike all the world over, I mutter to myself,
as Sefu explains this—full, on the one hand, of vanity, on the other, of
spite. The song refers to a poor girl appearing in borrowed finery,
who is satirized by her companions. In the next song it is my turn to
furnish the moral.
“Ignole yangala yangala meme mtuleke tuwakuhiyoloka.”
The sense appears to be about the following:—
“You are here assembled (for the unyago), rejoice and be merry.
We who have come here, we do not want to play with you, only to
look on.”
If Sefu is right, as there is every reason to suppose, these words are
to be understood as spoken by myself, they are either dictated by my
own delicacy of feeling: “I have no wish to intrude”—or they are
intended as a captatio benevolentiæ: “Please stay at a distance, white
man, or we shall be afraid!”
In spite of my discreet attitude, the performers do not seem to feel
quite easy, for they now sing till they grow tired:—
“Nidoba ho, nidoba ho.” (“It is difficult, it is difficult, truly.”)
This is followed by a long pause.
The second division of the programme goes on to repeat part of the
first. Still more completely muffled in their brightly-coloured cloths,
so that neither face nor arms are to be seen, the five girls come
forward as before, and march round to the right, the rest of the
company following them in the same order as previously. Now the
drums, which in the meantime have been tuned afresh over a
tremendous fire, strike up again, and the chorus starts: “Chihakatu
cha Kuliwile,” etc., with dance as before. This lasts fully half-an-
hour, and then the long file breaks up; the oldest of the instructresses
comes forward into the open space in front of the crowd, puts on a
critical expression, and waits for what is about to happen. This is not
long in showing itself. Like a gorgeous butterfly, one of the coloured
calico bundles separates itself from the mass, and trips gracefully
before the old woman, while the chorus bursts into song:—
“Nande è è, nande è è.”
The astonished white man, looking on, can only see clearly the
head and feet of the bundle, which are comparatively at rest—
everything between these extremities being an undistinguishable
blur. On boldly approaching, I make out that the girl is vibrating her
waist and hips, throwing herself to and fro with such velocity that the
eye cannot follow the lines of her figure. The performer retires after a
time, and the others follow, each in her turn, receiving praise or
censure from the high authorities convoked for the occasion. But not
even Sefu can tell me what the words of the song mean.
The third part follows. As full of expectant curiosity as myself, the
five young girls certified as having arrived at maturity are now gazing
at the arena. They have freed themselves from their wrappings, and
seem to feel quite at home, with their mothers and aunts all standing
round them. Then, with a quick, tripping step, another bundle of
cloth comes out of the bush, followed by a second, and, after a short
interval by a third and fourth. The four masks—for such, when they
turn round, they are seen to be—stand up two and two, each pair
facing the other, and begin the same series of movements which I
had already watched at Chingulungulu, comprising the most varied
manœuvres with arms and legs, contortions of the body above the
waist, quivering vibrations of the region below the waist. In short,
everything is African, quite authentic and primitive. I had seen all
these evolutions before, but was all the more struck with the whole
get-up of these strange figures. Makonde masks are now to be found
in the most important ethnographic museums, but no one, it
appears, has ever seen them in use—or, if so, they have not been
described. The masks are of wood, two of them representing men,
and two women. This is evident a hundred paces off, from the
prominence given to the pelele, whose white stands out with great
effect from the rigid black surface. The costume of the male and
female figures is in other respects alike, following the principle of
letting no part of the human form be seen—everything is swathed in
cloth, from the closely-wrapped neck to the tips of the fingers and
toes. This excessive amount of covering indicates the aim of the
whole—the masks are intended to terrify. It is young men who are
thus disguised; they do not wish to be recognized, and are supposed
to give the girls a good fright before their entrance on adult life. The
masks themselves in the first instance serve this purpose in a general
way, but their effect is still further heightened by making them
represent well-known bugbears: portraits of famous and much
dreaded warriors or robbers, heads of monstrous beasts, or, lastly,
shetani—the devil.[44] This personage appears with long horns and a
large beard, and is really terrible to behold.
MAKONDE MASKS
While the four masks are still moving about the arena—sometimes
all together facing each other, sometimes separating and dancing
round in a circle with all sorts of gambols—a new figure appears on
the stage. A tapping sound is heard as it jerks its way forward—
uncanny, gigantic; a huge length of cloth flutters in the morning
breeze; long, spectral arms, draped with cloth so as to look like
wings, beat the air like the sails of a windmill; a rigid face grins at us
like a death’s head; and the whole is supported on poles, a yard or
more in length, like fleshless legs. The little girls are now really
frightened, and even my bodyguard seem to feel somewhat creepy.
The European investigator cannot allow himself to give way to such
sensations: he has to gaze, to observe, and to snapshot.
The use of stilts is not very common in any part of the world.
Except in Europe they are, so far as I know, only used in the culture-
area of Eastern Asia, and (curiously enough) in the Marquesas
Islands (Eastern Pacific), and in some parts of the West Coast of
Africa. Under these circumstances, I cannot at present suggest any
explanation of their presence on the isolated Makonde plateau. Have
they been introduced? and, if so, from whence? Or are they a survival
of very ancient usages once prevalent from Cape Lopez, in the west to
this spot in the east, preserved at the two extremities of the area,
while the intervening tribes advanced beyond the old dancing-
appliances? My mind involuntarily occupies itself with such
questions, though, properly speaking, this is not the time for them,
as there are still many things to see.
That the stilt-dancer’s intention is to terrify, is evident from his
movements, quite apart from his disguise. In a few gigantic strides he
has reached the other side of the fairly spacious arena, and drives the
natives squatting there back in headlong flight; for it looks as if the
monster were about to catch them, or tread them under foot. But it
has already turned away, and is stalking up to the five novices at the
other end: they, and others near them, turn away shrieking. Now he
comes within range of my camera—a click of the shutter, and I have
him safe. I could almost have imagined that I saw the man’s face of
consternation behind his mask—he stopped with such a start,
hesitated a moment, and then strode swiftly away.
This dancing on stilts can scarcely be a pleasure. The man is now
leaning, tired out, against the roof of one of the huts, and looks on
while the four masks come forward again to take part in the dance.
But the proceedings seem inclined to hang fire—the sun has by this
time climbed to the zenith, and the stifling heat weighs us all down.
A great many of the women taking part in the
ceremony have already dispersed, and those
still present are visibly longing for the piles of
ugali at home. I take down the apparatus and
give the word to start, and once more we are
forcing our way through the thorny thickets of
the Makonde bush towards Newala.
MAKONDE STILT-
DANCER. FROM A
DRAWING BY OMARI,
A MBONDEI
THE NJOROWE DANCE AT NEWALA
The indefatigable Sefu only allowed me one day in which to digest

the impressions of Niuchi, before announcing another important
expedition. Sefu lives only some thirty or forty yards away from us, in
a house built Coast-fashion. He is not, like Nakaam and Matola, a
native of the country, but has been transferred here from the coast as
an official of the German Administration, while the other two might
be compared with large landowners placed in a similar position on
account of their local standing and influence among the people. He
has rather more notion of comfort than is usual among his
congeners, for he has had very neat bamboo seats—some even with
backs to them, an unheard of luxury in this region—put up in his
baraza, where he holds shauris and also receives, with great dignity,
the leaders of passing caravans. Sefu spends all his spare minutes
with us; he arrives first thing in the morning, and shivers through the
evening with us in that temple of the winds which goes by the name
of the rest-house, and which we shall be compelled to close in with a
wall in order to get some protection against the evening gales.
Sefu, then, had a grand plan to propose. This time, he said, he
could show us a ceremony of the Wamatambwe at the village of
Mangupa. It was again a girls’ chiputu, that is, the conclusion of the
first course of instruction which these children of between eight and
eleven had been going through for some months in a special hut. But
the Matambwe procedure is in some points different from that of the
Yaos and the Makua; and, also, it was not far. If we started next
morning at 7.30, we should be in time to see the beginning after a
walk of an hour and a half.
I was able to form a slight idea of the famous Makonde bush on the
expedition to Niuchi—but it was very far from being an adequate
one. Much has been written about this form of vegetation, but I
believe the theme is inexhaustible. Not that this bush is remarkable
for æsthetic charms, for beautiful scenery, or abundance and variety
of vegetation. It is a perfectly uniform, compact mass of thin stems,
branches, leaves and tendrils. This is the unpleasant part of it; this
indescribably thick tangle lets no one pass unless he has first cut his
painful and toilsome way with axe and bill-hook. Our native troops
have gone through unspeakable sufferings in this way, in the last ten
years alone, especially in the war against Machemba. Things have
been made easier for us—the victorious struggle against the formerly
unreliable and often rebellious tribes of the south has led to the wise
measure of connecting every place of the slightest importance with

PDF Recommender System For Improving Customer Loyalty Katarzyna Tarnowska Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

PDF Recommender System For Improving Customer Loyalty Katarzyna Tarnowska Ebook Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PDF Recommender System For Improving Customer Loyalty Katarzyna Tarnowska Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

Recommender System for Improving

Customer Loyalty Katarzyna Tarnowska

Customer Loyalty and Supply Chain Management Business

Marketing and Customer Loyalty The Extra Step Approach

The Opt Out Effect Marketing Strategies that Empower

Fashion Recommender Systems Nima Dokoohaki

Customer Accounting: Creating Value with Customer

Managing Customer Relationships Using Customer Care

Recommender System with Machine Learning and Artificial

Group Recommender Systems An Introduction 1st Edition

More information about this series at http://www.springer.com/series/11970

ISSN 2197-6503 ISSN 2197-6511 (electronic)

Library of Congress Control Number: 2019932683

© Springer Nature Switzerland AG 2020

San Jose, USA Katarzyna Tarnowska

9.4.3 Sequence Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Fig. 2.1 The concept of Net Promoter Score as a way to quantify

Fig. 5.6 Initial manual preprocessing of the data before loading

Fig. 9.4 Analysis of the decision attribute—Customer

Table 5.1 Sample meta-actions influence matrix . . . . . . . . . . . . . . . . . .. 55

Table 8.4 Improving sentiment analysis further for the combined

1.1 Why Customer Experience Matters More Now?

© Springer Nature Switzerland AG 2020 1

1.2 Top (and Bottom) Line Reasons for Better Customer

1.3 What is Next?

1.4 Final Observations

Customer experience is a very robust element in the strategic arsenal of a company.

effective and powerful ways to improve customer experience. The recommender

2.1 Introduction to the Problem Area

Nowadays most businesses, whether small-, medium-sized or enterprise-level orga-

Fig. 2.1 The concept of Net

2.2 Dataset Description

2.3 Decision Problem

2.4 Problem Area

2.4.1 Attribute Analysis

• Where did we go wrong? What could be improved?

2.4.2 Attribute Reduction

2.4.3 Customer Satisfaction Analysis and Recognition

2.4.4 Providing Recommendations

In this chapter, different types of available information technology solutions support-

3.1 Customer Satisfaction Software Tools

3.2 Customer Relationship Management Systems

CRM is described as “managerial efforts to manage business interactions with cus-

3.3 Decision Support Systems

A DSS is an interactive computer-based system designed to help in decision making

3.4 Recommender Systems

Most recommender systems were applied in e-commerce settings, supporting cus-

And what is chiputu? It is the celebration of a girl’s arrival at

GIRLS’ UNYAGO AT THE MAKONDE HAMLET OF NIUCHI

PARTICIPANTS ASSEMBLING AT THE UNYAGO HUT

PRESENTATION OF CALICO BY THE MOTHERS

ARRIVAL OF THE NOVICES

GIRLS’ UNYAGO AT THE MATAMBWE VILLAGE OF MANGUPA. I

OLD WOMEN GROUPED ROUND THE GIRLS TO BE

DANCE OF THE INITIATES BEFORE THE OLD WOMEN

DEPARTURE OF THE INITIATES

GIRLS’ UNYAGO AT THE MATAMBWE VILLAGE OF MANGUPA.

I expected more satisfactory results from Medula; but the

Newala, middle of September, 1906.