Download the Full Version of textbook for Fast Typing at textbookfull.

com

Targeting Using Augmented Data in Database

Marketing Decision Factors for Evaluating External
Sources 1st Edition Bettina Hüttenrauch

https://textbookfull.com/product/targeting-using-augmented-
data-in-database-marketing-decision-factors-for-evaluating-
external-sources-1st-edition-bettina-huttenrauch/

OR CLICK BUTTON

DOWNLOAD NOW

Download More textbook Instantly Today - Get Yours Now at textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Data Mashup with Microsoft Excel Using Power Query and M:

Finding, Transforming, and Loading Data from External
Sources Adam Aspin
https://textbookfull.com/product/data-mashup-with-microsoft-excel-
using-power-query-and-m-finding-transforming-and-loading-data-from-
external-sources-adam-aspin/
textboxfull.com

Marketing Database Analytics Transforming Data for

Competitive Advantage 1st Edition Andrew D. Banasiewicz

https://textbookfull.com/product/marketing-database-analytics-
transforming-data-for-competitive-advantage-1st-edition-andrew-d-
banasiewicz/
textboxfull.com

Data Processing Handbook for Complex Biological Data

Sources 1st Edition Gauri Misra

https://textbookfull.com/product/data-processing-handbook-for-complex-
biological-data-sources-1st-edition-gauri-misra/

textboxfull.com

Big Data Analytics Using Multiple Criteria Decision-Making

Models 1st Edition Ramakrishnan Ramanathan

https://textbookfull.com/product/big-data-analytics-using-multiple-
criteria-decision-making-models-1st-edition-ramakrishnan-ramanathan/

textboxfull.com
NoSQL database for storage and retrieval of data in cloud
1st Edition Deka

https://textbookfull.com/product/nosql-database-for-storage-and-
retrieval-of-data-in-cloud-1st-edition-deka/

textboxfull.com

Marketing New Realities An Introduction to Virtual Reality

Augmented Reality Marketing Branding Communications 1st
Edition Hackl Cathy
https://textbookfull.com/product/marketing-new-realities-an-
introduction-to-virtual-reality-augmented-reality-marketing-branding-
communications-1st-edition-hackl-cathy/
textboxfull.com

Using Open Data to Detect Organized Crime Threats Factors

Driving Future Crime 1st Edition Henrik Legind Larsen

https://textbookfull.com/product/using-open-data-to-detect-organized-
crime-threats-factors-driving-future-crime-1st-edition-henrik-legind-
larsen/
textboxfull.com

Homicide Data Sources: An Interdisciplinary Overview for

Researchers 1st Edition Adam Dobrin (Auth.)

https://textbookfull.com/product/homicide-data-sources-an-
interdisciplinary-overview-for-researchers-1st-edition-adam-dobrin-
auth/
textboxfull.com

Building a Columnar Database on RAMCloud: Database Design

for the Low-Latency Enabled Data Center 1st Edition
Christian Tinnefeld
https://textbookfull.com/product/building-a-columnar-database-on-
ramcloud-database-design-for-the-low-latency-enabled-data-center-1st-
edition-christian-tinnefeld/
textboxfull.com
Bettina Hüttenrauch

Targeting Using
Augmented Data in
Database Marketing
Decision Factors for Evaluating
External Sources
Targeting Using Augmented Data
in Database Marketing
Bettina Hüttenrauch

Targeting Using
Augmented Data in
Database Marketing
Decision Factors for Evaluating
External Sources
Bettina Hüttenrauch, geb. Krämer
Frankfurt, Germany

Dissertation Johannes Gutenberg-Universität Mainz, 2015

OnlinePlus material to this book can be available on

http://www.springer-gabler.de/978-3-658-14577-4
ISBN 978-3-658-14576-7 ISBN 978-3-658-14577-4 (eBook)
DOI 10.1007/978-3-658-14577-4

Library of Congress Control Number: 2016942900

Springer Gabler
© Springer Fachmedien Wiesbaden 2016
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.

Printed on acid-free paper

This Springer Gabler imprint is published by Springer Nature

The registered company is Springer Fachmedien Wiesbaden GmbH
Preface

This dissertation is the result of my studies on data augmentation in

database marketing. I am particularly interested in data and how it can
be used to differentiate people in order to automatically generate individ-
ualized communication. I am always surprised on how much data we leave
behind as digital footprints and – at the same time – how little we know
about our customers as database marketing analysts. It is my utmost con-
cern to reduce this information gap.
I have concentrated on the combination of statistics and communication
during my media management degree at the University of Mainz. After an
internship at Lufthansa German Airlines, I was granted the possibility to
conduct my first data augmentation for the database marketing department
at Miles & More, the frequent flyer program of Lufthansa. I was employed
as a database marketing analyst in 2010 and have been working in this
exciting field since.
There are enterprises specialized in data collecting and usage, e.g.
Google, Facebook, and Apple. When using their platforms, we experience
targeted communication. However, we seldom really notice it, because rele-
vant information is not something one realizes (unless struck as daunting).
Much more often, we perceive advertisements as irrelevant, misplaced, or
inappropriate. The vast amount of companies does not have a detailed data
basis to segment and select customers for differentiated marketing commu-
nication.

V
 At work, we had many discussions about which external sources can be
used for database marketing purposes and how. There is a general reluc-
tance regarding for example volunteer surveys and social media sources. In
these sources, data has been collected for a non-representative subgroup
of customers only. Furthermore, most of the data cannot or must not be
matched on an exact basis. Data augmentation projects require a consid-
erable amount of know-how, time, and money. It is not approached, unless
the return on marketing investment can be anticipated.
To me, these external sources provide a set of opportunities. Their
contained information cannot be obtained otherwise. External sources are
beneficial, because they are up-to-date, easy to acquire, and cheap. Ser-
vice providers offering data fusion services are often overpriced and less
experienced in the respective data. Without much effort, internal database
marketing analysts can do a better job. By highlighting important facts to
regard, I want to facilitate the use of data augmentation in companies.
Also, I would like to encourage the academic discussion regarding data
augmentation in database marketing. A wide range of augmentation ap-
proaches has evolved, both in direct marketing and online marketing. How-
ever, the scientific foundation for these approaches is sparse. I believe that
the methods and use cases for data augmentation will advance, if the aca-
demic discussion is pushed. I would like to make a contribution to this
matter.
I am much obliged to thank all persons that supported me during my
dissertation project. First and foremost I thank Heinz-Werner Nienstedt
for his supervision, support, discussions, and enthusiasm. I was very happy
to be able to continue to study at your chair at the Johannes Gutenberg
University of Mainz.
At the same time, I worked at Lufthansa Miles & More as a database
marketing analyst. I could not have had a better working environment for
completing a dissertation, while working at the same time. I thank my
superiors for the positive acceptance of my ”hobby”, the flexibility, and the

VI
support in all matters – including the possibility to use SAS. I also thank
my team members for inspiring discussions and challenging questions. It
was good to be able to get a direct feedback from the future beneficiaries of
my work.
It was a great pleasure to get the chance to discuss my work with two
of the most important researchers I cited, Susanne Rässler and Pascal van
Hattum. Both of them surprised me by being eager to engage in my ideas
and plans. Thank you for inviting me to Bamberg and Amersfoort, respec-
tively, and for taking the time to review my work. I was was able to present
my working paper at the 2013 EMAC Doctoral Colloquium in Istanbul to
Arnaud de Bruyn, Arvind Rangaswamy, and Ujwal Kayande. Thank you
for pushing me and for being so motivating. We had a very productive
and encouraging atmosphere in our track. I also thank Heather Skinner
and Nick Lee for reviewing my work at the Academy of Marketing 20th
Doctoral Colloquium in Cardiff.
Finally, I would like to thank my husband, family and friends for moti-
vating and understanding me in any possible way. Thank you for familiar-
izing with my complex topic, for proof-reading, and for listening. You have
always been so proud and never questioned my success. I could not have
asked for any more support.

VII
Contents

1 Introduction to data augmentation in marketing 1

1.1 Research problem and relevance . . . . . . . . . . . . . . . . . 2
1.1.1 Irrelevant communication and unused chances . . . . . 2
1.1.2 Academic and practical research context . . . . . . . . 6
1.1.3 Research question and desired contribution . . . . . . 8
1.2 Research concepts and context . . . . . . . . . . . . . . . . . 9
1.2.1 Definition of data augmentation and terminology . . . 10
1.2.2 Specific characteristics of external information . . . . 14
1.2.3 Independence of source and target variables . . . . . . 17
1.2.4 Delineation of the research subject . . . . . . . . . . . 19
1.3 Research approach . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3.1 Case study with simulated sampled sources . . . . . . 26
1.3.2 Suitability of the case study method . . . . . . . . . . 28
1.4 Structure of the paper . . . . . . . . . . . . . . . . . . . . . . 36

2 Strategic motivation for data augmentation 41

2.1 Internal conditions within the company . . . . . . . . . . . . 42
2.1.1 Customer focus as a major marketing goal . . . . . . . 42
2.1.2 Targeting in marketing practice . . . . . . . . . . . . . 44
2.1.3 The marketing measure: conversion . . . . . . . . . . 48
2.2 External conditions around the company . . . . . . . . . . . . 49
2.2.1 Economic environment . . . . . . . . . . . . . . . . . . 50

IX
 2.2.2 Technological environment . . . . . . . . . . . . . . . . 51
2.2.3 Legal environment . . . . . . . . . . . . . . . . . . . . 52
2.2.4 Sociological and psychological environment . . . . . . 55
2.3 Sources for data augmentation . . . . . . . . . . . . . . . . . 56
2.3.1 Public and official sources . . . . . . . . . . . . . . . . 57
2.3.2 Company-owned sources . . . . . . . . . . . . . . . . . 59
2.3.3 Accumulated sources . . . . . . . . . . . . . . . . . . . 61
2.3.4 Comparison of sources . . . . . . . . . . . . . . . . . . 64
2.4 Implications for data augmentation . . . . . . . . . . . . . . . 67

3 Literature review on data augmentation 71

3.1 Evolution of data augmentation studies . . . . . . . . . . . . 72
3.1.1 Missing data problems and imputation . . . . . . . . . 73
3.1.2 Statistical matching . . . . . . . . . . . . . . . . . . . 78
3.1.3 Data augmentation in marketing . . . . . . . . . . . . 79
3.2 Process of data augmentation . . . . . . . . . . . . . . . . . . 81
3.2.1 Data screening . . . . . . . . . . . . . . . . . . . . . . 81
3.2.2 Data preparation . . . . . . . . . . . . . . . . . . . . . 86
3.2.3 Choice of the best data augmentation method . . . . . 90
3.2.4 Execution and internal evaluation of results . . . . . . 98
3.2.5 External evaluation of augmentation results . . . . . . 102

4 Methodological framework for data augmentation 105

4.1 Data augmentation specifics in marketing . . . . . . . . . . . 106
4.1.1 The customer database as recipient unit . . . . . . . . 107
4.1.2 Possible donor units and their characteristics . . . . . 108
4.1.3 Variable scales, values, and validity . . . . . . . . . . . 111
4.1.4 Conditional independence of source and target variables113
4.1.5 Source data mechanism of the donor unit . . . . . . . 117
4.2 Data augmentation model . . . . . . . . . . . . . . . . . . . . 119
4.2.1 Populations and samples . . . . . . . . . . . . . . . . . 120
4.2.2 Variables . . . . . . . . . . . . . . . . . . . . . . . . . 123

X
 4.2.3 Univariate pattern approach . . . . . . . . . . . . . . . 127
4.2.4 Data augmentation target values and uncertainty . . . 129
4.2.5 Ignorability of the source data mechanism . . . . . . . 133

5 Test design for evaluating the source characteristics 139

5.1 Research question conceptualization . . . . . . . . . . . . . . 140
5.1.1 Conceptual model . . . . . . . . . . . . . . . . . . . . 140
5.1.2 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . 145
5.2 Dataset generation and case study set up . . . . . . . . . . . 150
5.2.1 Population characteristics and test design . . . . . . . 150
5.2.2 Choice of variables . . . . . . . . . . . . . . . . . . . . 154
5.2.3 Test samples . . . . . . . . . . . . . . . . . . . . . . . 166
5.2.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 172

6 Analysis of data augmentation KPIs 175

6.1 Preserved descriptive data and measures . . . . . . . . . . . . 176
6.2 Quality checks . . . . . . . . . . . . . . . . . . . . . . . . . . 179
6.2.1 Predictive power and discrimination . . . . . . . . . . 179
6.2.2 Managerial implications . . . . . . . . . . . . . . . . . 183
6.3 Accuracy and precision of results . . . . . . . . . . . . . . . . 184
6.3.1 Total correct classification rate . . . . . . . . . . . . . 185
6.3.2 Correct classification rate of the target value . . . . . 193
6.4 Model lifts assessing the results . . . . . . . . . . . . . . . . . 197
6.4.1 Model lift (chance) . . . . . . . . . . . . . . . . . . . . 198
6.4.2 Model lift (target) . . . . . . . . . . . . . . . . . . . . 207
6.4.3 Model lift (uniform) . . . . . . . . . . . . . . . . . . . 211
6.4.4 Model lift (source) . . . . . . . . . . . . . . . . . . . . 215
6.4.5 Comparison of the KPIs . . . . . . . . . . . . . . . . . 219
6.5 Conversion probability lift . . . . . . . . . . . . . . . . . . . . 222

7 Analysis of results and test of hypotheses 233

7.1 Validating the source data mechanism . . . . . . . . . . . . . 234

XI
 7.1.1 Test based conditional dependency calculation . . . . 235
7.1.2 Model based conditional dependency calculation . . . 241
7.1.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 245
7.2 Influences of source characteristics . . . . . . . . . . . . . . . 248
7.2.1 Representation . . . . . . . . . . . . . . . . . . . . . . 248
7.2.2 Overlap . . . . . . . . . . . . . . . . . . . . . . . . . . 250
7.2.3 Number of donors and surplus . . . . . . . . . . . . . 253
7.3 Overall estimation of influences . . . . . . . . . . . . . . . . . 256
7.3.1 Regression models estimating the model lift . . . . . . 258
7.3.2 Estimating the best method to be chosen . . . . . . . 262
7.3.3 Conversion probability and conclusion . . . . . . . . . 266
7.4 Managerial implications . . . . . . . . . . . . . . . . . . . . . 272

8 Limitations of data augmentation and outlook 279

8.1 Limitations and alternatives . . . . . . . . . . . . . . . . . . . 280
8.2 Further research opportunities . . . . . . . . . . . . . . . . . 284
8.2.1 Ignorability of the source data mechanism . . . . . . . 285
8.2.2 Deeper exploration of the augmentation process . . . . 286
8.2.3 Further augmentation opportunities and use cases . . 287
8.2.4 Uplift models . . . . . . . . . . . . . . . . . . . . . . . 289

9 Summary 291

Appendix 303

Glossary 315

References 319

Curriculum Vitae 341

Abstract 343

XII
List of Figures

1.1 Schematic illustration of data augmentation . . . . . . . . . . . . 11

3.1 Examples of missing data patterns . . . . . . . . . . . . . . . . . 75

4.1 Source data mechanism in the data augmentation context . . . . 118

4.2 Notations used in the data augmentation model . . . . . . . . . . 120
4.3 Data augmentation sources by size and overlap . . . . . . . . . . 122
4.4 Dimensions of the overall population . . . . . . . . . . . . . . . . 124
4.5 Dimensions of donor unit and recipient unit . . . . . . . . . . . . 125
4.6 Augmented recipient unit . . . . . . . . . . . . . . . . . . . . . . 132

5.1 Conceptual model . . . . . . . . . . . . . . . . . . . . . . . . . . 142

5.2 Deviations from the optimum influencing the model lift . . . . . 146
5.3 Population pyramid of the case study data . . . . . . . . . . . . 151
5.4 Design of the data augmentation case study . . . . . . . . . . . . 153
5.5 Link variables and distributions available from the dataset . . . . 155
5.6 Target variables and distributions available from the dataset . . 161
5.7 Auxiliary variables and distributions . . . . . . . . . . . . . . . . 165
5.8 Sources created for the case study . . . . . . . . . . . . . . . . . 167
5.9 Sources by overlap, size, and representation . . . . . . . . . . . . 170

6.1 Quality check for number of reproduced target values . . . . . . 182

6.2 Distribution of T CCRmodel by k . . . . . . . . . . . . . . . . . . 188

XIII
6.3 Distribution of T CCRmodel by k and IQV . . . . . . . . . . . . 190
6.4 Distribution of T CCRmodel by method . . . . . . . . . . . . . . . 192
6.5 Distribution of CCRtarget by ntarget . . . . . . . . . . . . . . . . 195
6.6 Distribution of CCRtarget by method . . . . . . . . . . . . . . . 196
6.7 Distribution of M Lchance by k . . . . . . . . . . . . . . . . . . . 203
6.8 Distribution of M Lchance by k and IQV . . . . . . . . . . . . . . 204
6.9 Distribution of M Lchance by method . . . . . . . . . . . . . . . . 205
6.10 Distribution of M Ltarget by ntarget . . . . . . . . . . . . . . . . . 209
6.11 Distribution of M Ltarget by method . . . . . . . . . . . . . . . . 210
6.12 Distribution of M Lunif orm by method . . . . . . . . . . . . . . . 214
6.13 Exemplary schematic illustration of a M Lsource calculation . . . 216
6.14 Distribution of M Lsource by method . . . . . . . . . . . . . . . . 219
6.15 Conversion probability measures . . . . . . . . . . . . . . . . . . 224
6.16 Conversion probability for a 87% overlap rate source . . . . . . . 227
6.17 Conversion probability for a 8% overlap rate source . . . . . . . . 229
6.18 Scatter plot for uplift by method and optimal selection . . . . . . 230
6.19 Distribution of CLglobal by method . . . . . . . . . . . . . . . . . 232

7.1 Distribution of M Lchance by source data mechanism . . . . . . . 240

7.2 Distribution of M Lchance by categories . . . . . . . . . . . . . . . 244
7.3 Distribution of M Lchance for representative sources . . . . . . . . 249
7.4 Distribution of M Lchance by o and method . . . . . . . . . . . . 251
7.5 Distribution of M Lchance by d and method . . . . . . . . . . . . 254
7.6 Distribution of M Lchance by surplus and method . . . . . . . . . 256
7.7 Best method per case as measured by M Lchance . . . . . . . . . 257
7.8 Effect plots for best method for o and k . . . . . . . . . . . . . . 264
7.9 Distribution of CLglobal by best method . . . . . . . . . . . . . . 267
7.10 Distribution of CLglobal by source data mechanism . . . . . . . . 270
7.11 Practical guide for evaluating external sources . . . . . . . . . . . 276

9.1 Fit diagnostics for nearest neighbor hot deck . . . . . . . . . . . 312

9.2 Fit diagnostics for logistic regression . . . . . . . . . . . . . . . . 313

XIV
List of Tables

1.1 Data augmentation terminology . . . . . . . . . . . . . . . . . . . 13

1.2 Alternative study designs for answering the research question . . 30

2.1 Rating of available data augmentation sources . . . . . . . . . . . 65

2.2 SWOT analysis implying strategic suggestions . . . . . . . . . . 68

3.1 Target variables using conditional mode imputation . . . . . . . 92

4.1 Association table of source and a special interest . . . . . . . . . 114

4.2 Conditional association table of source and a special interest . . 115

5.1 Examples of nominal associations between link variables . . . . . 157

5.2 Examples of ordinal associations between link variables . . . . . 158
5.3 Overview of associations between link variables . . . . . . . . . . 159
5.4 Results of check for predictive power . . . . . . . . . . . . . . . . 163

6.1 Descriptive information saved for every augmentation . . . . . . 177

6.2 Exemplary quality check for Target 1 and Donors c s6 . . . . . . 181
6.3 Two-way frequency table for calculating a T CCR . . . . . . . . . 186
6.4 Exemplary T CCRmodel values . . . . . . . . . . . . . . . . . . . 187
6.5 Exemplary CCRtarget (y4 ) values . . . . . . . . . . . . . . . . . . 194
6.6 Upper and lower boundaries for T CCRchance and M Lchance . . . 201
6.7 Exemplary M Lchance values for Target 1 and Donors c s6 . . . . 202

XV
6.8 Exemplary M Ltarget values for Target 1 and Donors c s6 . . . . 208
6.9 Frequency table for a uniformly distributed target variable . . . 213
6.10 Exemplary M Lunif orm values for Target 1 and Donors c s6 . . . 213
6.11 Distributions of Target 1 in recipient and donor unit . . . . . . . 217
6.12 Exemplary M Lsource values for Target 1 and Donors c s6 . . . . 218
6.13 Comparison of different model lift measures . . . . . . . . . . . . 220
6.14 Exemplary CLglobal and CMglobal values . . . . . . . . . . . . . . 228

7.1 Example for test based conditional independence calculation . . 238

7.2 Test based calculations of conditional dependencies . . . . . . . . 239
7.3 Model based calculations of conditional dependencies . . . . . . . 243
7.4 Comparison of test and model based results . . . . . . . . . . . . 246
7.5 Source data mechanisms and augmentations results . . . . . . . . 247
7.6 Regression parameters for nearest neighbor hot deck . . . . . . . 259
7.7 Regression parameters for logistic regression . . . . . . . . . . . . 261
7.8 Parameters for the best method to be used . . . . . . . . . . . . 263
7.9 Final measures for data augmentation results . . . . . . . . . . . 271

9.1 Number of link variable classes by numbers of observations . . . 303

9.2 Distribution of CCRtarget by k . . . . . . . . . . . . . . . . . . . 304
9.3 Parameter estimates for M Lchance by method . . . . . . . . . . . 304
9.4 Distribution of M Ltarget by k . . . . . . . . . . . . . . . . . . . . 305
9.5 Parameter estimates for M Ltarget by method . . . . . . . . . . . 305
9.6 Parameter estimates for M Lunif orm by method . . . . . . . . . . 305
9.7 Parameter estimates for M Lsource by method . . . . . . . . . . . 305
9.8 Parameter estimates for CLglobal by method . . . . . . . . . . . . 306
9.9 Par. estimates for M Lchance by source data mechanism . . . . . 306
9.10 Par. estimates for M Lchance by model based calculations . . . . 307
9.11 Parameter estimates for M Lchance by o and method . . . . . . . 308
9.12 Distribution of M Lchance by o and method (o > 0) . . . . . . . . 308
9.13 Parameter estimates for M Lchance by d (Conditional mode) . . . 309
9.14 Parameter estimates for M Lchance by d (Logistic regression) . . 310

XVI
9.15 Parameter estimates for M Lchance by d (Nearest neighbor) . . . 311
9.16 Par. estimates for M Lchance using nearest neighbor hot deck . . 311
9.17 Parameter estimates for M Lchance using logistic regression . . . 314

XVII
List of Abbreviations

ANOVA analysis of variance

B2B business-to-business

B2C business-to-consumer

BDSG Bundesdatenschutzgesetz

C2C consumer-to-consumer

CMH Cochran-Mantel-Haenszel

CPL conversion probability lift

CRM customer relationship management

DWH data warehouse

IP internet protocol

KPIs key performance indicators

MAR missing at random

MCAR missing completely at random

MNAR missing not at random

ROMI return on marketing investment

XIX
Sh.-W. Shapiro-Wilk test

SWOT analysis of strength, weaknesses, opportunities, and threads

TMG Telemediengesetz

UWG Gesetz gegen den unlauteren Wettbewerb

XX
Definition of Variables

A areas under and between the conversion probability curves

a number of auxiliary variables

b absolute value

c customer indicator variable

CCR correct classification rate of the target value

CL conversion probability lift

CM conversion probability lift magnitude

D donor unit in the data augmentation model

d number of donors

d number of representing units

F F statistic as derived from the F distribution

h hit indicator variable

IQV index of qualitative variation

j number of values per link variable

k number of values per target variable

XXI
l number of link variables

M L model lift

N number of observations

n number of selected customers for a direct marketing campaign

O overlap between recipient unit and donor unit

o number of overlapping elements

P overall population in the data augmentation model

p overall number of elements

R recipient unit in the data augmentation model

r number of customers or recipients

S set of source data mechanism indicator variables

t number of target variables

T CCR total correct classification rate

v number of link variable classes

W W statistic for Shapiro-Wilk’s test for normally distributed residuals

X set of link variables

Y set of target variables

Z set of auxiliary variables

XXII
Chapter 1

Introduction to data augmentation in

marketing

Although marketing specialists spend a considerable amount of time, money,

and know-how on relevant marketing campaigns, everybody is confronted
with more less than well personalized advertisements every day. Relevance
in this context is defined by attracting the positive attention of the recipient
to the content or offer. While it can take weeks, if not months, to create
these campaigns, the customers receiving the offer only need seconds to
decide whether an offer is relevant or not. Especially in direct marketing,
where prospective customers are purposely selected, nothing is less desirable
to a marketer than an offer being ignored. The right selection and allocation
of marketing communication is an every-day database marketing problem.
The data available on customers is not always sufficient to adequately
define target groups and to meet the marketing goals. At the same time,
external information is available encompassing many relevant facts. By
augmenting this data, both companies and customers would profit from
the increased relevance. While the information overload would decrease for
the customers, wastage could be avoided from a company’s point of view.
Which external information sources are suitable for data augmentation and

1
© Springer Fachmedien Wiesbaden 2016
B. Hüttenrauch, Targeting Using Augmented Data in Database Marketing,
DOI 10.1007/978-3-658-14577-4_1
how they can be used are the central questions of this study. This question
has not yet been regarded in academic research so far.
In this chapter, we lay the basis for our study by explicating the prob-
lem of irrelevant communication and unused possibilities and deriving our
research question. We explain important concepts and the context of data
augmentation in database marketing in order to determine the research field
and applicability. Eventually, we describe the research approach with which
we answer the research question so as to establish practical guidelines for
how to assess external sources upfront regarding their suitability for data
augmentation in database marketing.

1.1 Research problem and relevance

Data augmentation can increase marketing efficiency. Database marketing
analysts are responsible for finding the right target groups for individualized
and personalized marketing communication. But the available information
in the customer database is limited, so that augmenting data has become
a valuable alternative to directly collecting data from customers. In this
chapter, we explicate the problem of irrelevant communication and the un-
used possibilities from external sources in order to motivate the research
question. We describe the academic and practical research context and the
current state of research regarding the topic. From the practical need, the
chances given by external sources, and the lack of attention in academic
discussion our research question and desired contribution are derived.

1.1.1 Irrelevant communication and unused chances

Direct marketing has the goal of maximizing the profit of individual cus-
tomers by increasing their spending volumes, exploiting their willingness
to pay, and reducing their communication costs to a minimum, so as to
grow the return on marketing investment (ROMI). Ideally, customers re-

2
ceive only relevant information to increase their interest, satisfaction and
eventually their loyalty to the company. The need for efficient marketing is
derived from the economic environment as described in chapter 2.2.1. The
profit related to a customer centric communication approach (in contrast
to a product centric approach) has already been recognized in the 20’s cen-
tury. It has been promoted both in practice and in academics (e.g. by Huldi
(2002), Link and Hildebrand (1993), Rossi, McCulloch, and Allenby (1996),
and Schweiger and Wilde (1993)). The customer focus as a major market-
ing goal is described in more detail in chapter 2.1.1. The vision of customer
relationship management (CRM) is to convey the right information to the
right person at the right location and time.
Data is the basis for all direct marketing activities. In order to best reach
the customers, a lot of information is required on their preferences, needs
and wants, and state in the customer life cycle. The better the available
data, the more precise target group selections can be made. Database mar-
keting structures in companies are available, extending the mere collection
of transaction data to more sophisticated data mining methods and models
(e.g. Adriaans and Zantinge (1998), Küsters (2001), Ratner (2001b), and
Weiss and Indurkhya (1998)). The data is stored in a customer database,
usually a customer data warehouse (DWH). These structures are explicated
in more detail in chapter 4.1.1.
With the existing structures and processes, relevant communication
should be very easy to deliver. But all too often, one is negatively sur-
prised on how little companies know. For example when being female, aged
20-30, and living in a metropolitan area, one most likely receives online ads
from dating websites offering handsome bachelors in the respective area. It
fits the (few) available data, but might not be relevant. There are several
reasons why companies lack relevant information.
Targeting for direct mailings, online marketing, or newer media is done
semi-manually or implemented for automatic deployment by database mar-
keting analysts. They use the information available in order to select the

3
right target groups for campaigns and promotions. When information is
not ready at hand, database marketing analysts make assumptions, build
models, and derive predictions in order to target the right customers. If a
retailer wants to promote a luxury product, e.g. an expensive watch, it does
not have the variable ”affinity for expensive watches” in its database. But
it has monthly spending, transaction volumes, and comparable products
bought. The retailer would target customers with a suitable transaction
history for his campaign. However, the affinity for other high end prod-
ucts and expensive watches do not have to be correlated and may lead to
irrelevant communication.
Even if detailed information is available, it is most commonly available
only for a small, highly active portion of the customers. The big portion
of occasional and inactive customers is not well describable by sufficient
criteria. Consequently, well targeted promotions are possible only for a
small group of customers, which is not sufficient for sales purposes. All
other customers receive standardized offers. The resulting wastage is high.
Other information may be available and useful, but may not be allowed
to be used. Online behavioral data such as surfing behavior, mailing aware-
ness, and click frequency could help to identify customers generally affine for
ads and commercial information. But unless customers have not been asked
for their permission to use this behavioral data, companies must not apply
it for targeting on an individual basis in Germany. The legal environment
for database marketing is described in more detail in chapter 2.2.3.
Additional knowledge about customers is often available in aggregated
form only. Surveys are conducted frequently to gain deeper insight into
customer segments. For example, grandmothers may love buying presents
for their kindergarten grandchildren, or singles working in the finance sector
may be prone to take last minute offers for their vacations. But variables
like family information, relationship status, or employment are usually not
available from the customer database. Thus, although these segments are

4
well describable and campaign actions are clear, it is not possible to identify
these segments in order to treat them individually.
The result of these problems is a disadvantage for both customers and
companies. If the data is not the right data, cannot be interpreted, must not
be used, or is not sufficient in order to differentiate target groups, relevance
degenerates to simple gender-age-region-schemas as in the example. Because
of these limitations, ad space (available media in terms of channels and
platforms) cannot be used efficiently. The ad burden for customers is high,
and so are costs per contact when comparing contacts with sales. This is
particularly true for below the line media, thus all channels through which
customers can directly and individually be reached and where access is
limited, e.g. email, letter, SMS, or promoters at the point of sale.
Data augmentation can be the answer to many of the problems de-
scribed. Supplemental variables not available for the customers can be
matched to individual profiles based on link variables present in both the
customer database and the external source. They are derived from external
sources; e.g. the company website, a customer survey, market media studies,
or social media applications. As a result, definite variables can be used for
targeting, rather than demographic target group descriptions, derivations,
or common knowledge. Values are augmented for all customers, so that not
only very active customers can be differentiated. Because data is augmented
by groups rather than on a personal level, also sensitive information, e.g. a
personal income level, is addable. Even aggregated data can be used, if an
appropriate augmentation set up with suitable link variables, is chosen.
Today, we face a situation in which data is collected at various touch
points, but little information is actually used to improve the marketing
communication. A list of available external sources and their possibilities is
given in chapter 2.3. With the exponential growth of data and information,
CRM is experiencing a renaissance in academics and management. While
the possibility and necessity of the use of external information for CRM

5
has already been recognized (Arnold, 2011; Breur, 2011), a detailed study
describing the process and which sources to use is still missing.
Many external sources have disadvantageous features, like being incom-
plete, partially overlapping with the customer group, not representative,
or generally small. To apply the contained target variables to the cus-
tomer database can lead to biased results. One reason why external sources
are not used for database marketing purposes is the anxiety that available
sources are not valuable in terms of data utility, meaning that it cannot
be assessed upfront whether data augmentation results will be reliable and
effective. With our study, we assess different forms of sources in order to
give practical insights on which sources to use and how.

1.1.2 Academic and practical research context

The conditions for data augmentation with external sources in database
marketing relevant today have been established in different research fields.
In statistics and market research, methods for deriving joint information on
people from different sources have been developed. The usage of augmenta-
tion techniques in marketing has begun with the introduction of widely-used
electronic tracking systems and the change to a customer focus. Along with
the emergence of social media marketing, local and mobile marketing, the
merging of external data and the usage in direct marketing have come to
form a new research field. These branches of science converge at data aug-
mentation with external sources in database marketing.
This work ties in with the research projects of Kamakura and Wedel
(1997, 2000), Putten et al. (2002a; 2002b), and Gilula, McCulloch, and
Rossi (2006). They used data augmentation, or data fusion, in the context
of marketing. Especially, it contributes to Hattum and Hoijtink’s (2008b,
2008a) idea of data fusion and extends it to a broader range of possible
applications. The two comprehensive works of Rässler (2002) and D’Orazio,
Di Zio, and Scanu (2006) build the methodological basis for this extension

6
of the data augmentation idea. Fundamental ideas therein, in the studies
named previously, and in this work are attributed to Rubin (Rubin, 1976;
Little & Rubin, 2002), who fathomed the conditions for data augmentation
in the context of missing data theory. This work enlarges the field of data
augmentation and statistical matching with a new and practical focus on
using existing sources for data augmentation in database marketing.
The literature on data augmentation has been developed in statisti-
cal and marketing journals, before it found its way into specialized jour-
nals like Database Marketing & Customer Strategy Management and the
Journal of Targeting, Measurement and Analysis for Marketing, which
have been consolidated into the Journal of Marketing Analytics as of 2013
(Palgrave Macmillan, 2013). Major studies from a statistical point of view
can be found in the Journal of the American Statistical Association and
Biometrika. The literature from a marketing perspective is more disperse,
including the Journal of Marketing Research, the International Journal of
Market Research, and the Journal of Direct, Data and Digital Marketing
Practice.
Data augmentation in database marketing is a practice oriented field
of research. Database marketing analysts will directly benefit from the
solutions. They work in the marketing or analytics department of com-
panies, or for specialized agencies offering data augmentation services, like
tns infratest (2012) in Germany or The SmartAgent Company (2013) in
the Netherlands. The professional field is constantly growing, and so is
the need for tools and methods. While companies like Google, Facebook,
and Apple are expected to know a considerable amount of information on
any arbitrary person, companies whose main business is not data collection
have difficulties finding manageable ways of handling external data. Little is
published on data augmentation for practitioners. There is no professional
or academic exchange of ideas and approaches for data augmentation.
Data augmentation for database marketing is promoted at many places
and with detailed descriptions of internal and external sources – often with-

7
out going into detail about the processes and challenges related to it (Breur,
2011; Kuhner, 2013; Schiff, 2010). This study examines the contemporary
situation for data augmentation with external sources in database market-
ing. It theoretically analyzes the special features relevant for data augmen-
tation in this field and takes into consideration the challenges related to the
using external sources. It is a starting point for a professional and academic
conversation regarding the topic, facilitating a much more standardized and
sophisticated development of the research field.

1.1.3 Research question and desired contribution

The question arising from the problems encountered in database market-
ing practice can be answered by the information available from external
sources. But under which circumstances can external sources be used for
data augmentation? It is necessary that database marketing analysts gain
information from the data augmentation results, but not sufficient. In a
marketing context, the information itself is only a mediator for the target-
ing goal, which is to increase conversion probabilities. Thus, the following
research question is derived.

Which external data augmentation sources are able to increase

conversion probabilities?

The question inherently suggests that it is possible to increase conver-

sion probabilities by augmenting external data. Previous works by Putten
et al. (2002a) and Hattum and Hoijtink (2008b) suggested that data aug-
mentation is able to increase conversion rates, at least under certain cir-
cumstances. However, it has never been considered what is special about
data augmentation using external sources in database marketing and un-
der which circumstances data augmentation results significantly increase
conversion probabilities. In particular, it has never been assessed com-
prehensively which sources are suitable for data augmentation in database
marketing and which source characteristics are essential in this assessment.

8
 Most of the literature on data augmentation explicitly or implicitly refers
to representative sources. These are convenient, but their features cannot
be generalized for all forms of sources available today. A theoretical con-
tribution of this study is a comprehensive description of sources and their
formal characteristics. The quality of data augmentation depends on which
link and target variables are used and on the predictive power of the link
variables regarding a target variable. The augmentation methods also have
an influence on the augmentation results.
The managerial contribution of this study is to establish a guide for aug-
menting external data in database marketing. It defines which characteris-
tics are relevant for assessing external sources and the variables contained.
The guide provides information on how to choose the right augmentation
method and on how to manage expectations regarding augmentation re-
sults. Database marketing analysts are not familiar with using the new
sources for data augmentation. By establishing a practical guide for ex
ante evaluating data augmentation sources, the idea of data augmentation
becomes more tangible. This study provides a starting point for broader
usage. Once a process has been established, the techniques can be further
refined by practitioners and applied to many different cases.

1.2 Research concepts and context

The research objective of this study is to examine data augmentation with
external sources in database marketing for establishing guidelines regard-
ing their usage in practice. In order to answer the research question and
to facilitate the understanding for our approach, important concepts and
the context of data augmentation in database marketing are explained in
this chapter. The exact terminology and definition of data augmentation
is given, as well as illustrative examples for available external sources. The
different sources available in today’s digital world have disadvantageous
characteristics, such as being incomplete, not representative, or generally

9
Another Random Scribd Document
with Unrelated Content
 · · · ·

There saw she wade

In the heavy streams
Men—foul perjurers,
And murderers.

· · · ·

Brothers slay brothers:

Sisters’ children
Shed each other’s blood.
Hard is the world;
Sensual sin grows huge.
There are sword-ages, axe-ages,
Earth-cleaving cold;
Storm-ages, murder-ages,
Till the world falls dead,
And men no longer spare
Or pity one another.

· · · ·

Mimer’s sons play,

But the world is kindled
By the ancient
Gjallarhorn.
Loud blows Heimdall,
His sound is in the air:
Odin talks
With the head of Mimer.

Quivers then Yggdrasil,

The strong-rooted ash:
Rustles the old tree
When Jötun gives way.
All things tremble
In the realms of Hel,
Till Surtur’s son
Swallows up Odin.

Garmer he shouts
By the Gnipa-hall
The band must burst
And the wolf fly.

Hrymer drives eastward,

Bears his shield before him;
 Jormungand welters
In giant fierceness.
The waves thunder;
The eagles scream;
Death rends the corpses
And Nagelfar gives way.

Köl hies eastward;

Come must Muspel’s
Folk to the sea.
Loke rows afar;
All the children of madness
Follow the wolf,
Bileist’s brother
Journeyeth with them.

Surtur fares southward.

With flickering flames
From his sword
God’s sun flashes.
Break the stone mountains;
The weird women flee,
Men throng Hel’s dread roads,
And Heaven is rent.”

Then Surtur flings fire over the world.

“The sun grows dark.
Earth sinks in the sea.
From heaven vanish
The lustrous stars.
High from the flames
Rolls the reek;
High play the fires
’Gainst heaven itself.

· · · ·

Up, sees she come

Yet once more,
The earth from the sea,
Gloriously green.

· · · ·

Then comes the Mighty One

To the great Judgment—
The great above all—
 He who guides all things.
Judgments he utters;
Strifes he appeaseth;
Laws He ordaineth
To flourish for ever.

· · · ·

In Gimlé the lofty

There shall the hosts
Of the virtuous dwell,
And through all ages
Taste of deep gladness.”

FROM THE “SÓLAR LJÓD”

The “Sólar Ljód”—“Sol” or “Sun-song”—was composed by Sæmund himself,
the collector of the Edda, and a Christian priest, ages before the time of
Dante.
“By the Nornors’ seat
Nine days I sate,
Then to horse was lifted.
The sun of the giant race
Gleamed sadly
Out of heaven’s weeping clouds.

Without and within

Seemed I to journey
Through the seven worlds
Above and below.
Better path I sought
Than there was to find.

And now to be told is

What first I beheld
In the home of torture.
Scorched birds were flying—
Wretched souls in myriads,
Thick as mosquito legions.

Flying saw I
Hope’s dragons
And fall in drear waste places.
They shook their wings
Till to me seemed that
Heaven and earth were rent.

The stag of the sun

Southward saw I journey.
His feet stood
On earth, but his huge antlers
Traversed the heavens above him.

Northward saw I ride

The sons of the races;
Seven they were together.
From the full horn they drank
The purest mead
From wells of heavenly strength.

The winds stood still,

The waters ceased to flow.
Then heard I a dread cry.
There for their husbands
False vengeful women
Ground earth for food.

Bloody stones
Those women dark
Dragged sorrowfully,
Their gory hearts
Hung from their breasts
Weighed with heavy weights.

Many men
Along the burning ways
Sore wounded saw I go.
Their visages
Seemed deeply dyed
With blood in murder shed.

Many men
Saw I amongst the dead
Without one hope of grace.
Pagan stars there stood
Over their heads
All scored with cruel runes.

Men saw I too

Who enviously had scowled
Upon the good of others.
Bloody runes
Were on their breasts
Ploughed out by hands of men.

Men saw I there

All full of woe,
All mazed in wondering.
This do they win
Who to eternal loss
Love this world only.

Men saw I too

Who sought always to snatch
From others their possessions.
In throngs they were,
And to the miser’s hell
Bore groaning loads of lead.

Men saw I next

Who many had bereaved
Of life and goods,
And through the hearts of these
For ever fiercely ran
Strong venom snakes.

Men too I saw

Who never would observe
Sabbaths and holy days.
Their unblessed hands
Fast rivetted together
With ever burning stones.

Men too I saw

Who with huge brag and boast
On earth did vaunt themselves.
Here their clothes
Were vilely squalid
And with fire enwrapt.

Men saw I too

Who with their slanderous breath
Had blasted others.
Hel’s ravens
Remorselessly their eyes
Tore from their heads.

But all the horrors

Thou canst not know
Which Hel’s condemned endure.
Sweet sins
There bitterly are punished,
False pleasures reap true pain.

· · · ·

Men did I see

Who the Lord’s laws
Had followed stanchly.
Purest light
For ever growing clearer
Passed brightly o’er their heads.

Men did I see

Who with unwearied zeal
Did seek the good of others.
Angels read
The holy books
Upon their radiant heads.

Men did I see

Who with sharp fasts
Their bodies had subdued;
God’s holy hosts
Before them all bow’d down
And paid them highest homage.

Men did I see

Who had their mothers
Piously cherished,
And their place of rest
Amid heaven’s beams
Shone gloriously.

Holy maids there were

Who their pure souls
Had kept unsoiled by sin,
And souls of those
Who their rebellious flesh
Did ever sternly quail.

Lofty chariots saw I

Travel through heaven
Having access to God;
And they were filled with those
Who causelessly
Had on the earth been slain.
 Father Almighty!
Illustrious Son!
And Holy Spirit of Heaven.
Thee do I implore,
Who didst make all things,
To keep us from all sin!”

Sæmund concludes this remarkable poem with these strophes:

“This song
Which I have taught thee
Thou shalt sing unto the living.
The Sun’s song,
Which in its solemn theme
Hath little that is feigned.

Here do we part,
But part again to meet
On the Great Day of men.
Oh, my Lord!
Give the dead rest,
Comfort to those who live!

Wonderful wisdom,
To thee in dream is sung,
’Tis truth which thou hast seen!
And no man is so wise
Of all who are created
As, ere this, to have heard
One word of this Sun’s Song!”

FROM THE POEMS RELATING TO SIGURD &

BRYNHILD.
From the heroic poems relating to Sigurd and Brynhild—the originals from
which the German “Niebelungen-lied” is taken—the following passage is
extracted. In it Gudrun, in conversation with Thjodreck, describes her youth
before the murder of Sigurd: “A maid was I amongst maidens; my mother
reared me lovely in bower. Well loved I my brothers, till me Gjuké apportioned
with gold, with gold apportioned and gave me to Sigurd. So raised himself,
Sigurd, over the sons of Gjuké, as the green lily above the grass grows; or the
high-antlered stag, above other beasts; or the fire-red gold above the silver
grey. My brothers were incensed that I should have a husband more illustrious
than any. Sleep they could not, nor decide on anything, before they Sigurd had
caused to perish. Grangé (Sigurd’s steed) galloped to the Ting (assembly of the
people), wild was his neighing, but Sigurd himself was not there. All the horses
were covered with sweat, and with blood of the contenders.
“Weeping I went to speak to Grané, the blood sprinkled; of his master I asked
him; then hung down Grané mournfully his head, for the creature knew that
his lord was not living. Long did I wander, long was I confused in mind before
of the Prince I could ask after my King.”
The “Hávamál”—“Odin’s High Song”—displays a shrewd insight into human
nature, and contains many maxims, both of a moral and social kind, which one
would scarcely expect to find embodied in the heathen ethics of an ancient
Scandinavian Scald. The whole poem is here presented to the reader.

HÁVAMÁL.
I.

“In every corner

Carefully look thou
Ere forth thou goest;
For insecure
Is the house when an enemy
Sitteth therein.

II.

Hail him who giveth!

Enters a guest.
Where shall he be seated?
Yet, ill shall fare he
Who seeks his welfare
In other men’s houses.

III.

Fire will be needful

For him who enters
With his knees frozen.
Of meat and clothing
Stands he in need
Who journeys o’er mountains.
 IV.

Water is needful,
A towel and kindness
For this guest’s welcome;
Kind inclinations
Let him experience;
Answer his questions.

V.

Good sense is needful

To the far traveller;
Each place seems home to him.
He is a laughing-stock
Who, knowing nothing,
Sits mid the wise.

VI.

With the deep thinker

Speak thou but little;
But guard well thy temper;
When the noble and silent
Come to thy dwelling,
Least errs the cautious.

VII.

Good sense is needful

To the far traveller;
Least errs the cautious;
For a friend trustier
Than good understanding
Findeth man never.

VIII.

A cautious guest
When he comes to his hostel
Speaketh but little;
With his ears he listeneth;
With his eyes he looketh;
Thus the wise learneth.

IX.

Happy is he
Who for himself winneth
Honour and friends.
All is uncertain,
Which a man holdeth
In the heart of another.

X.

Happy is he
Who prudent guidance
From himself winneth;
For evil counsel
Man oft receiveth
From the breast of another.

XI.

No better burden
Bears a man on his journey
Than mickle wisdom.
Better is she than gold
Where he is a stranger;
In need she is a helper.

XII.

No better burden
Bears a man on his journey
Than mickle wisdom.
No worse provision
Takes a man on his journey
Than frequent drunkenness.

XIII.

Ale is not so good

As people have boasted
For the children of men.
For less and still less,
As more he drinketh,
Knows man himself.

XIV.

The hern of forgetfulness

Sits on the drunkard,
And steals the man’s senses.
By the bird’s pinions,
Fettered I lay,
In Gunlada’s dwelling.

XV.

Drunken I lay,
Lay thoroughly drunken,
With Fjalar the wise.
This is the best of drink,
That every one afterwards
Comes to his senses.

XVI.

Be silent and diligent,

Son of a Prince,
And daring in combat;
Cheerful and generous,
Let every man be,
Till death approaches.

XVII.

A foolish man fancies

He shall live for ever
If he shuns combat.
But old age will give
To him no quarter,
Although the spear may.

XVIII.

The fool stares about

When he goes on a visit,
Talks nonsense or slumbers.
All goes well
When he can drink,
For then the man speaks his mind.

XIX.

He, he only
Who has far travelled,
Has far and wide travelled,
Knoweth every
Temper of man,
If he himself is wise.
 XX.

If cups thou lackest

Yet drink thou by measure:
Speak what is seemly or be still.
No one will charge thee
With evil, if early
Thou goest to slumber.

XXI.

The gluttonous man,

Though he may not know it,
Eats his life’s sorrow:
Lust of drink, often
Makes the fool, foolish
When he comes mid the prudent.

XXII.

The flocks they have knowledge

When to turn homeward
And leave the green pastures;
But he who is foolish
Knoweth no measure,
No bounds to his craving.

XXIII.

An evil man
And a carping temper
Jeer at all things.
He knows not;
He ought to know,
That himself is not faultless.

XXIV.

A foolish man
Lies awake the night through
And resolves on many things.
Thus is he weary
When the day cometh;
The old care remaineth.

XXV.

A foolish man
Thinks all are friendly
Who meet him with smiles;
But few he findeth
Who will aid his cause,
When to the Ting he cometh.

XXVI.

A foolish man
Thinks all are friendly
Who meet him with smiles.
Nor knows he the difference
Though they laugh him to scorn
When he sits ’mong the knowing ones.

XXVII.

A foolish man
Thinks he knows everything
While he needs not the knowledge.
But he knows not
How to make answer
When he is questioned.

XXVIII.

A foolish man,
When he comes into company
Had better keep silence.
No one remarketh
How little he knows
Till he begins talking.

XXIX.

He appears wise
Who can ask questions
And give replies.
Ever conceal then
The failings of others,
The children of men.

XXX.

Who cannot keep silence

Uttereth many
A word without purport.
The tongue of the garrulous,
Which keepeth back nothing,
Talks its own mischief.

XXXI.

Hold in derision
No one, although he
Come as a stranger.
Many a one, when he has had
Rest and dry clothing,
Thou mayest find to be wise.

XXXII.

He seemeth wise
Who in speech triumphs
O’er mocking guests.
The talkative man
Knows not at the table
If he talks with his enemies.

XXXIII.

Many are friendly

One to another;
Yet storm ariseth.
Strife will arise
For ever, if one guest
Affronteth another.

XXXIV.

Thou mayst dine early

Unless thou art going
Unto the banquet.
Sits he and flatters;
Hungry he seemeth,
Yet few things he learneth.

XXXV.

Long is the journey

To a deceitful friend
Though he dwell near thee.
But, direct lies the path
To a friend faithful,
Though he dwelleth afar off.
 XXXVI.

Do not too frequently

Unto the same place
Go as a guest.
Sweet becomes sour
When a man often sits
At other men’s tables.

XXXVII.

One good house is there

Though it be humble:
Each man is master at home.
Though a man own but
Two goats and a straw-rick,
’Tis better than begging.

XXXVIII.

One good house is there

Though it be humble:
Each man is master at home.
The man’s heart bleedeth
At every mealtime
Who his food beggeth.

XXXIX.

Without his weapon

Goes no man
A-foot in the field.
For it is unsafe
Out on the by-paths
When weapons are needful.

XL.

Never found I so generous,

So hospitable a man
As to be above taking gifts.
Nor one of his money
So little regardful
But that it vexed him to lend.

XLI.

He who has laid up

Treasures of wealth
Finds want hard to bear.
Adversity often uses
What was meant for prosperity,
For many things are contrary to expectation.

XLII.

With weapons and garments,

As best may be fitting
Give thou thy friends pleasure.
By gifts interchanged
Is friendship made surest;
If the heart proffers them.

XLIII.

Let a man towards his friend,

Ever be friendly,
And with gifts make return for gifts.
With thy cheerful friend
Be thou cheerful;
With thy guileful friend on thy guard.

XLIV.

Let a man towards his friend

Ever be friendly;
Towards him and his friend.
But with an enemy’s friend
Can no man
Be friendly.

XLV.

If thou hast a friend

Whom thou canst confide in
And wouldst have joy of his friendship,
Then, mingle thy thoughts with his,
Give gifts freely,
And often be with him.

XLVI.

If thou hast another,

Whom thou hast no faith in
Yet wouldst have joy of his friendship,
Thou must speak smoothly;
Thou must think warily,
And with cunning pay back his guile.

XLVII.

Yet one word

About him thou mistrusteth
And in whom thou hast no reliance.
Thou must speak mildly,
More so than thou meanest;
Paying back like with like.

XLVIII.

Young was I formerly;

Then alone went I,
Taking wrong ways.
Rich seemed I to myself
When I found a companion;
For man is man’s pleasure.

XLIX.

The noble, the gentle

Live happiest,
And seldom meet sorrow.
But the foolish man,
He is suspicious,
And a niggard grieves to give.

L.

I hung my garments
On the two wooden men
Who stand on the wall.
Heroes they seemed to be
When they were clothed!
The unclad are despised.

LI.

The tree withereth

Which stands in the court-yard
Without shelter of bark or of leaf.
So is a man
Destitute of friends.
Why should he still live on?
 LII.

Even as fire,
Burns peace between enemies,
For the space of five days.
But on the seventh
It is extinguished,
And the less is their friendship.

LIII.

Only a little
Will a man give;
He often gets praise for a little.
With half a loaf
And a full bottle
I won a companion.

LIV.

Small are the sand-grains,

Small are the water-drops:
Small human thoughts:
Yet are not these
Each of them equal.
Every century bears but one man.

LV.

Good understanding
Ought all to possess,—
But not too much wisdom.
Those human beings
Whose lives are the brightest,
Know much and know it well.

LVI.

Good understanding
Ought all to possess,
But not too much knowledge.
For the heart of a wise man
Seldom is gladdened
By knowledge of all things.

LVII.

Good understanding
Ought all to possess,
But not too much knowledge.
Let no one beforehand
Inquire his own fortune.
The gladdest heart knoweth it not.

LVIII.

Brand with brand burneth

Till it is burned out:
Fire is kindled by fire.
A man among men
Is known by his speech;
A fool by his arrogance.

LIX.

Betimes must he rise

Who another man’s life
And goods will obtain.
The sleeping wolf
Seldom gets bones.
No sluggard wins battle.

LX.

Betimes must he rise

And look after his people
Who has but few workmen.
Much he neglecteth
Who sleeps in the morning.
On the master’s presence depends half the profit.

LXI.

Like to dried faggots,

And hoarded up birch bark,
Are the thoughts of a man,
The substance of firewood
May last, it is true,
A year and a day.

LXII.

Cleanly and decent,

Ride men to the Ting
Although unadorned.
For his shoes and apparel
Nobody blushes,
Nor yet for his horse, though none of the best.

LXIII.

Question and answer

Is a clever thing,
And so it is reckoned.
To one person trust thyself,
Not to a second.
The world knows what is known unto three.

LXIV.

Bewilderedly gazes
On the wild sea, the eagle,
When he reaches the strand.
So is it with the man
Who in a crowd standeth
When he has but few friends there.

LXV.

Every wise man

And prudent, his power will use
With moderation.
For he will find
When he comes ’mong the brave
That none can do all things.

LXVI.

Let every man

Be prudent and circumspect
And cautious in friendship.
Often that word
Which we trust to another
Very dear costs us.

LXVII.

Greatly too early

Came I to some places;
Too late to others.
Here the feast was over;
There unprepared.
Seldom opportunely comes an unwelcome guest.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com