Contributions published at Marketing and Market Research (René Algesheimer)

null

null

One of the most challenging problems in technological forecasting is to identify as early as possible those technologies that have the potential to lead to radical changes in our society. In this paper, we use the US patent citation network (1926-2010) to test our ability to early identify a list of historically significant patents through citation network analysis. We show that in order to e?ectively uncover these patents shortly after they are issued, we need to go beyond raw citation counts and take into account both the citation network topology and temporal information. In particular, an age-normalized measure of patent centrality, called rescaled PageRank, allows us to identify the significant patents earlier than citation count and PageRank score.

With customer lifetime value (CLV) playing a great role for firms to evaluate marketing decisions, this bachelor thesis aims at giving an overview of how accurate the estimation of individual CLV is by analyzing the purchase history of multiple retailers. In a first step, findings from previous studies are summarized and different dataset characteristics are discussed. In a second step, the results of customer behavior prediction with a hiatus heuristic and with an advanced analytics are compared. Moreover, it will compare the impact of variations of the prediction period over multiple cohorts by analyzing real-world data from multiple retailers. The results of the customer base analysis indicate that the predictive performance strongly depends on the dataset characteristics and the length of the prediction period. Keywords: customer lifetime value, Pareto/NBD, hiatus heuristic, forecasting, cohort comparison, customer base analysis

During the last decades, complex social, economical and biological systems have been studied using agent-based models (ABM). ABM are a powerful tool to discover analytic truths at the macroscopic-level when simple rules at the agent-agent interaction level are assumed. Despite great achievements in discovering analytic truths in complex systems and a large number of large datasets. Statistical models aimed to discover factual truths in complex systems have not reached the rigorous approach of econometrics models. In this paper, we introduce a network model, power law random graph model (PRGM) formulated at the agent-agent interaction level via the concept of q-conditional independence where q can be interpreted as idiosyncratic correlations or an interaction term between well-defined social mechanisms. We show that the exponential random graph models (ERGM) are the subclass of PRGM with q = 1. Motivated by the derivation of ERGM via the Boltzmann-Shannon entropy by Park and Newman, we present a second formulation of the PRGM via Tsallis entropy. Next, we construct a subclass of PRGM, called q-Markov graph models, defined by simple dependency assumptions and that violates Gaussian approximation of the network statistics. The violation of Gaussian approximation is caused by competitive social mechanisms, and it enriches PRGM with distributions ranging from bimodal, skewed and flat. Our findings open the question What warrants Gaussian approximations used to justify factual evidence in complex systems? Finally, with the help of the subclass q-Bernoulli random graph models and using two networks datasets of friendships between students in classrooms in Switzerland and the US, we show how the idiosyncratic correlation q helps to address the problem of models placing too much probability mass around a few type of networks. Although the problem of placing too much probability is well documented in poor-fitting network models, we show that this problem is inherited from the exponential decay of rare events in ERGM, and it occurs in poor-fitting- as well as overfitting models.

During the last decades, complex social, economical and biological systems have been studied using agent-based models (ABM). ABM are a powerful tool to discover analytic truths at the macroscopic-level when simple rules at the agent-agent interaction level are assumed. Despite great achievements in discovering analytic truths in complex systems and a large number of large datasets. Statistical models aimed to discover factual truths in complex systems have not reached the rigorous approach of econometrics models. In this paper, we introduce a network model, power law random graph model (PRGM) formulated at the agent-agent interaction level via the concept of q-conditional independence where q can be interpreted as idiosyncratic correlations or an interaction term between well-defined social mechanisms. We show that the exponential random graph models (ERGM) are the subclass of PRGM with q = 1. Motivated by the derivation of ERGM via the Boltzmann-Shannon entropy by Park and Newman, we present a second formulation of the PRGM via Tsallis entropy. Next, we construct a subclass of PRGM, called q-Markov graph models, defined by simple dependency assumptions and that violates Gaussian approximation of the network statistics. The violation of Gaussian approximation is caused by competitive social mechanisms, and it enriches PRGM with distributions ranging from bimodal, skewed and flat. Our findings open the question What warrants Gaussian approximations used to justify factual evidence in complex systems? Finally, with the help of the subclass q-Bernoulli random graph models and using two networks datasets of friendships between students in classrooms in Switzerland and the US, we show how the idiosyncratic correlation q helps to address the problem of models placing too much probability mass around a few type of networks. Although the problem of placing too much probability is well documented in poor-fitting network models, we show that this problem is inherited from the exponential decay of rare events in ERGM, and it occurs in poor-fitting- as well as overfitting models.

null

Ziel dieser Arbeit ist es, den Zusammenhang zwischen Werten und Werte demonstrierendem Verhalten, Einstellungen sowie Eigenheiten aufzuzeigen um daraus Verhaltensprofile abzuleiten, welche Manager verwenden könnten um Mitarbeiter zu rekrutieren und Arbeitsteams zusammenzustellen. Mithilfe der Strukturgleichungsmodellierung (SGM) wurde, durch die Daten einer Umfrage in Polnischen und Schweizer Schulen ein Modell getestet, das die postulierten Hypothesen abbildet. Der Autor dieser Arbeit beschränkt sich bei der Überprüfung der Hypothesen auf die Schweizer Jugendlichen der vierten Welle. Die Ergebnisse bestätigen die positiven Korrelationen zu dem Werte demonstrierenden Verhalten. Die Korrelationen mit Einstellungen und Eigenheiten müssen von weiteren Studien untersucht werden, damit die Personalführung von den Resultaten profitieren kann. Manager legen grossen Wert darauf, dass ihre Mitarbeiter gute Leistungen erbringen. Deswegen wird in dieser Arbeit, neben den Zusammenhängen zwischen Werten und weiteren Variablen, untersucht, welche Werteausprägungen zu guter Leistung führen. Die Resultate zeigen, dass Personen, denen Werte wie Toleranz, Hilfsbereitschaft und Gerechtigkeit wichtig sind, höhere Leistungen erbringen. The aim of this thesis is to show the relationship between values and values of demonstrating behavior, attitudes, and peculiarities, in order to derive behavioral profiles that managers could use to recruit employees and form work teams. Using the structure equation modelling (SGM), a modell that displays the postulated hypothesis was tested, using data collected from Swiss and Polish schools. The author of this thesis is limited to examining the hypotheses of the Swiss young people of the fourth wave. The results confirm the positive correlations to the value-demonstrating behavior. The correlations to attitudes and peculiarities must be examined by further studies so that the personnel management can benefit from the results. Managers attach great importance to ensuring that their employees perform well. For this reason, in addition to the correlations between values and further variables, this work examines the values which lead to good performance. The results show that people which see tolerance, willigness to help and justice as important values, perform better.

Notwithstanding the usefulness of system dynamics in analysing complex policy problems, policy design is far from straightforward and in many instances trial-and-error driven. To address this challenge, we propose to combine system dynamics with network controllability, an emerging eld in network science, to facilitate the detection of e ective leverage points in system dynamics models and thus to support the design of in uential policies. We illustrate our approach by analysing a classic system dynamics model: the World Dynamics model. We show that it is enough to control only 53% of the variables to steer the entire system to an arbitrary nal state. We further rank all variables according to their importance in controlling the system and we validate our approach by showing that high ranked variables have a signi cantly larger impact on the system behaviour compared to low ranked variables.

Both researchers and managers have discussed human values and their associated behavioural consequences from a theoretical, as well as from a practical point of view. The consequences of having similar values were assessed in this thesis, suggesting that values might lead to more trust, greater satisfaction and higher performance, in organisations, schools, and other institutions. Schwartz’s value theory provided the framework for the analysis of the consequences. Using Schwartz’s Portrait Value Questionnaire (PVQ) and Picture Based Value Survey (PBVS), the author examined if people who trust each other have similar values. Moreover, the correlation of trust with satisfaction and of satisfaction with performance was investigated. The data were collected from 1’335 Swiss students from ages 8 to 17. The analysis indicated a tendency to trust people who hold similar values. Furthermore, the literary review of trust and satisfaction suggested a positive correlation between the two variables, although the influence could not be supported with the data used in this thesis. The author did not find any statistical significance of the impact of satisfaction on performance. However, the thesis provides noteworthy insights for institutions, showing the relevance and positive consequences associated with human values.

In line with customer centricity, analyzing customer lifetime value provides competitive advantage. Academics are exerted to develop and implement stochastic models for customer base analysis. A common approach is the Pareto/NBD model that predicts customer churn and transaction behavior simultaneously. In practice however, managerial decisions mostly follow simple heuristics. To implement stochastic models, their predominance in accurate predictions must be proven. This study compares the performance of the Pareto/NBD model to simple heuristics in determining active customers, predicting future transaction and spending levels, and identifying future top customers. Two real world data sets from retailers are analyzed with multiple cohorts over various prediction periods. The heuristics determine active customers more accurately than the Pareto/NBD model, while the latter excels in predicting future behavior. Furthermore, the length of the prediction period will affect the accuracy of the prediction. Hence, an explicit recommendation of any approach to managerial practice cannot be derived.

Valuing customers is essential to any firm and enables marketers to identify key customers. Customer lifetime value (CLV) is a central metric for valuing customers. It describes the long-term economic value of customers and gives managers an idea of how customers will evolve over time. To model CLVs in continuous non-contractual business settings, probabilistic customer attrition models such as the Pareto/NBD model are the preferred choice in literature and practice. Their ability to simultaneously forecast both, customer’s actual lifetime and future transactions is unique. However, empirical evidence suggests that standard probabilistic customer attrition models do not outperform basic management heuristics. A possible explanation is that standard probabilistic customer attrition models do not consider important contextual factors, such as direct marketing or regularity purchase patterns. Recently an implementation that allows the inclusion of such time-varying contextual factors for the continuous non-contractual setting has been proposed. In this study, we compare the predictive accuracy of this model extension, the standard Pareto/NBD model, and managerial heuristics based on three key metrics: (1) distinction of active and inactive customers, (2) forecasts of future purchase level and (3) aggregated purchase volume of the entire customer base. The comparison is carried out for multiple datasets and multiple prediction horizons. Our findings have strong implications for both, marketing practice and research. Besides giving detailed recommendations on when to use which modeling approach, we also provide practical advices for applying probabilistic customer attrition models.

null

null

Apart from many other functions and purposes they serve, social media can also be channels for corporate communication. In order to derive meaningful marketing implications from social media, companies need to know how their corresponding users of corporate social media channels are positioned in the so-called value space – and compare this positioning to the public image of their own organization. In this thesis, we investigate a sample of 30 large global financial institutions on Twitter, with respect to sentiments expressed in their users’ Tweets, as well as “Schwartz’s four value dimensions”, namely “Conservation”, “Openness to Change”, “Self-Enhancement” and “Self-Transcendence”. We identify the positions of these firms in the value space of their users’ minds and compare these positions to the ones these companies see themselves. Furthermore, we use a dynamic approach to investigate the sentiment- and values development. Finally, we compare the findings identified in Twitter conversations to the findings identified in news articles listed on Google News. Our contribution to the existing literature is twofold. First, we are introducing our framework in a previously under-researched context. While sentiment detection has been applied in many different contexts, we are matching the corresponding words used in Tweets with keywords listed in databases to be used for conventional sentiment detection in the context of financial institutions. Second, we are providing databases providing customized keywords for “Schwartz’s value dimensions” and functions to match these keywords with words used in Tweets. Our functions are intended to be tools for positioning users in a two-dimensional value space, which can identify corresponding positions of each company in their users’ minds. Our results presented within this thesis suggest that there exist some discrepancies between the positions in which organizations see themselves, and the positions in which consumers see the organizations.

null

Opinion leaders (OLs), or influencers, have been studied extensively since the concept was created in 1944. The boost in modern technology and the constant flux of people’s values and beliefs provide exciting new opportunities as well as challenges for researchers and managers in understanding how situations and contexts change the way people perceive social influence, and hence to optimise information or product diffusion process. Recognising the diversity of contexts and its importance to social influence perceptions, we’ve firstly tried to build up a comprehensive system that helps people identify and understand contexts, and subsequently find the most appropriate targets to promote behavioural change in either research or marketing activities. Through the extensive literature review on contextual influence, we’ve noticed social influence under uncertain circumstances to be an understudied topic. Therefore we’ve carefully designed a laboratory experiment intending to find out the answer to the question: are OLs still as influential in uncertain situations? We compare OLs to another very important type of influencer - peers - and have constructed a novel hypothesis that, when there’s not much risk, people do tend to follow the OL’s advocates; conversely, in ambiguous circumstances, peers are more influential than “opinion leaders”, even when the peer opinion comes from an anonymous aggregation of people who do not even share personal relationships with the decision maker. In our prior-to-experiment test we were able to prove the former part of this claim, and our lab experiment is designed to testify the latter. With sufficient resources, we should also be able to find out the extent to which different levels of risk and uncertainty affect the strength of peer and OL influence.

Smog disasters are becoming more and more frequent and may cause severe consequences on the environment and public health, especially in urban areas. Social media as a real-time urban data source has become an increasingly effective channel to observe people׳s reactions on smog-related health hazard. It can be used to capture possible smog-related public health disasters in its early stage. We then propose a predictive analytic approach that utilizes both social media and physical sensor data to forecast the next day smog-related health hazard. First, we model smog-related health hazards and smog severity through mining raw microblogging text and network information diffusion data. Second, we developed an artificial neural network (ANN)-based model to forecast smog-related health hazard with the current health hazard and smog severity observations. We evaluate the performance of the approach with other alternative machine learning methods. To the best of our knowledge, we are the first to integrate social media and physical sensor data for smog-related health hazard forecasting. The empirical findings can help researchers to better understand the non-linear relationships between the current smog observations and the next day health hazard. In addition, this forecasting approach can provide decision support for smog-related health hazard management through functions like early warning.

Background: he article presents the first assessment of the structure and hierarchy of values using the Schwartz theory in Vietnam. Given the near-universal prevalence of the structure of values, we expected this to be found in Vietnam as well. Regarding the hierarchy of values, we expected the hierarchies in the Vietnamese samples to be quite different from the pan-cultural baseline because of Vietnam’s traditional culture. Participants and procedure: e administered a Vietnamese version of the Portrait Value Questionnaire (PVQ-40) to adult respondents in three regions, Ho Chi Minh City/Saigon (n = 521), Hue (n = 538), and Hanoi (n = 533). Results: ultidimensional scaling (MDS) and confirmatory factor analysis (CFA) analyses of the total sample and the samples from each region supported the theorized circular structure. However, it was necessary to combine some adjacent values in the circle in each sample. The hierarchies of values in the samples differed substantially from the pan-cultural hierarchy identified by Schwartz and Bardi. The values exhibited partial scalar invariance across the three regional samples, justifying comparisons of means. Conclusions: We discuss the differences in value hierarchies among regions and between Vietnam and other countries by examining the cultural, historical, and social structural characteristics specific to Vietnam and its regions. In future research, it would be worthwhile to explore causes, processes and consequences of the values in Vietnam.