10.
doo, “doo – Smart Online Event Management & Marketing Software,” doo.net, 2016. [Online]. Available: https://doo.net. [Accessed 02 08 2016].
11.
D. Neumann und M. Liebmann, Cloud Software – Chance or Risk for Small Non-Profit Institutions?, Freiburg: University of Freiburg – Public and Non-profit Management, 2015.
© Springer-Verlag GmbH Germany 2018
Claudia Linnhoff-Popien, Ralf Schneider and Michael Zaddach (eds.)Digital Marketplaces Unleashedhttps://doi.org/10.1007/978-3-662-49275-8_59
59. Consumer Journey Analytics in the Context of Data Privacy and Ethics
Andreas Braun1 and Gemma Garriga2
(1)Allianz SE, Munich, Germany
(2)INRIA Research Center, Rocquencourt, France
Andreas Braun (Corresponding author)
Email: [email protected]
Gemma Garriga
Email: [email protected]
59.1 Introduction
Big Data Analytics plays a key role in transforming traditional businesses to digital ones. While many typical applications of Big Data are relatively well understood and uncritical—typically all analytics on non‐person related data and information—both the biggest immediate business impact as well as challenge comes with person related data (or personally identifiable information, PII). These threats to companies now become even more tangible as we are approaching the implementation of the General Data Protection Regulation/Act (GDPR/A) in Europe. In this paper, we focus on challenges implementing data protection, privacy, and ethics by the example of consumer analytics and digital consumer decision journeys more specifically. We argue that respecting privacy under the regime of the GDPR requires trusted technical means and measures, such as Privacy by Design (PbD) and Privacy Enhancing Technology (PET) frameworks.
The consumer decision journey [1–3] refers to the life cycle of a consumer from the moment of first contact to the ultimate purchase (see Fig. 59.1); this cycle includes also end‐to‐end user experience, starting from an initial trigger, via several touch points, to a purchasing decision, continue using the product or service, and finally being an advocate. The consumer decision journey is structured in loops, representing consideration, enjoyment, and loyalty (different authors may use other terms). The model of a consumer journey appreciates the fact that todays’ consumers have many options, and many of them may get into the picture during the evaluation phase, hence creating a far more dynamic and agile process than traditional funnels.
Fig. 59.1The life cycle in a consumer decision journey
In fact, the graph of a journey tells the overall story across multiple touch points in various channels. The big change in the digital world are especially the speed and vastly increasing turnaround times, the dynamic set of options, and the number of external influencers, such as product ratings, social media, etc. As today most touch points are digital (social media, page and ad impressions, TV (return channel), product ratings, search topics, …), the data representation of the digital consumer journey abounds.
Big Data and analytics capabilities have been in turn developed to analyze digital journeys and to understand and influence consumer journeys in real‐time. Potential questions are translated into hypothesis applied in analytics, such as e. g., Consumer profile—by which category can a distinct consumer be described? interest—what is the specific consumer interest? Next best product—what is she most likely to buy next? Cross‐ and upsell—what other product or services can be offered? Trends in consumer interest and behavior—how is consumer interest changing and how to adapt to stay relevant? How can a brand innovate based on trends in consumer behavior?
59.1.1 Big Data Enabled Digital Consumer Analytics
Big Data infrastructures and analytics are the technology enablers that can ingest and process large semi‐ and unstructured data needed to build models and predict propensity behavior. For example, when ingesting large digital consumer journey data sets, these are represented by heterogeneous set of digital traces such as web logs, cookies from different devices and pages, tracking pixels, sensor data, online surveys, discussions and online blogs, emails and others. With data engineering we merge and match all the various collected data sources to build knowledge representation from the sources. Machine Learning (ML) models can then utilize these representations indifferent ways: e. g. by learning behaviors from known historical data, by modeling higher level abstractions with complex structures, like non‐linear transformations, by finding complex patterns to explain and model a given phenomenon. Learned models can be used in real‐time to predict propensities (i. e., to convert, to churn, to cross and up‐sell, next best action etc.).
Hence, Big Data Analytics actively shapes the decision journeys of consumers by allowing businesses to continuously optimize processes, understand the needs of consumers and stay in control. Consumer journeys have become central to the customer’s experience and as important as the products itself. As in an ever faster digital game, customers switch digital services and products at the tip of a finger, user experience will be the key differentiator. In the Experience Economy [4], each product or service will need to strive for constant user ‘sensations’ in contrast to focusing company’s internal needs.1
59.1.2 A Broader Notion of Digital Marketing
As competition and the speed of decisions making—for or against a product—ever increases, today’s challenge for businesses is to stay in the relevant set of brands of consumers. This broadens the borders of sales and expands the notion of marketing. We believe that marketing will become programmatic blending into real‐time and online algorithms to go beyond recommending the next best product and eventually compiling product bundles and creating services on‐the‐fly tailored for one single customer and beyond ‘one‐size fits all’ (the segment of one).
Although understanding digital journeys is key to stay relevant, we argue that promoting journey innovation and customer experience is the next step. Businesses can use Big Data Analytics to understand and promote new digital business models for example. This implies going to phases of continuous prototyping and gathering data of how new services, features and totally new ideas are received by users. The analysis of the consumer and customer reactions tells the story of how these new business ideas can develop into bigger contexts and markets.
59.1.3 From Analytical Veracity, Data Privacy and Ethics to Trust in the Digital Age
Big Data is a technology‐driven game changer with impact on but not limited to traditional IT, analytics, marketing, and privacy. We now realize and get a firm grasp on the positive influencers, e. g., how horizontal‐scaling systems change IT, how ML takes analytics to the next level, and how digital consumer journeys broaden the notion and role of marketing, while marketing in turn boosts user experience. In contrast, we just start to understand the challenges. Big Data can and is used to unveil personally identifiable information (or PII) and identities in supposed anonymized information. For instance, mobile movement patterns and trajectories can be used to identify almost all individuals in a country, even though all source data is anonymized [5]. People are increasingly tracked, profiled, and analyzed in social media, web, while using mobile phones, or connected cars.
The following discussion is based on our experiences in conceiving, building, and anchoring a Big Data Analytics competence center in a very large multi‐national organization. We will illustrate our approach to focus on Big Data‐driven marketing, consumer analytics, and experience to deliver value fast. Most importantly, we describe how the privacy by design framework (general frameworks presented in [6, 7]) can help to ensure privacy and how we can address data ethics to deliver trust in the digital age.
59.2 Staying in the Relevant Set of Consumers
In today�
��s digital business world, consumers are empowered to compare and benchmark. The traditional linear sales funnel [1, 8]—starting with an initial set of options and working towards a decision—is not valid anymore. Consumers arrived at an interactive and iterative process. This brings fail fast and trial and error to the purchasing process and the digital consumer journey as such became agile 2. The combination of Big Data technology and advanced analytics helps businesses to enter and remain in the relevant set of their customers. This means that all internal processes are supported in an automated way to streamline the different journey steps.
Artificial Intelligence (AI) and Machine Learning (ML) play an important role to help businesses staying in the relevant set of consumers. ML is a discipline in AI and refers to technology and algorithms that can learn and make predictions based on large data sets, including historical events. For example, deep learning, is inspired by advancements in neurosciences where algorithms learn abstract representations from data by using multiple processing layers with complex structures. Deep learning is nowadays applied successfully e. g., in image and speech recognition. Another relevant example of ML is online learning, where algorithms adapt predictive models on the fly as data is being continuously processed. Notable examples can be found in online advertisement and customer journey classification.
The ML approach in Big Data Analytics is different from, and overcomes the limitations of, rule engines or statistics. Patterns and features learned can be enormously complex and go way beyond human limitations of cognitive perception. ML models can be designed to be dynamically, interactive, and online, which means that they can be self‐improving ‘on‐the‐fly’. ML today further allows for very fast and real‐time application in digital consumer journeys. Using ML and predictive analytics in general is not a new topic for digital businesses and brands that digitize their interactions with customers. Within the context of Big Data and the analysis of Digital Consumer Journeys, ML goes one step beyond by boosting prediction power and supporting the optimization and automatization of processes within the journey decision.
Examples of successful digital customer journey/analytics use cases developed within the insurance business are e. g.: Customer retention and improved customer experience. We build models to understand and segment consumers and prospect customers. With advanced data engineering, we blend data from all collected data sources—either based on past interactions, transactional systems, product usage, Web logs and internal cookies—and build ML models that will accurately predict the propensity of a customer to churn based on the stored digital traces. Beyond that, we also build complementary models that can identify the “pain points” that most likely will turn customers unsatisfied with the experience of our brand. For that, we collect also NPS (Net Promoter Score, [9]) data and correlate it to the current observed effects on churning customers. Our models can subsequently be used to monitor those issues and target the most efficient repair. Repairing the experience of a customer goes in hand by offering a better service and a better deal, for example for specific customer segments. In this sense, ML models will then predict what is the next best action for a customer in order to retain her and continue being in the relevant set. All these models are streamlined in real‐time systems that support re‐contacting the customer with the next best action.
Proactive personalization. We use information about users navigating on the Allianz websites—e. g. based on past interactions if we detect that the user is a customer, on past navigation patterns and external data sources—in order to dynamically customize the experience of a user on our website. Our models predict the most probable interest of the user at a given point in time and therefore, Web pages and applications can be re‐arranged accordingly. The required analytic models for proactive understanding of the customer needs, help to personalize and optimize next steps of a consumer journey. For example, for good value customers we can take the action to personalize with the best‐value offer; e. g. if a frequent traveler customer comes to the Allianz website, we can personalize content in order to help the customer understand the most useful dynamic upgrades of her travel insurance.
Consumer quote conversion and best discount accommodation. We identify consumers with high potential and interest to become future customers of Allianz. Based on the digital traces of previously submitted quotes of a given insurance product, we build models that predict the best moment to approach the consumer and the best possible discount to funnel the buying decision. We also integrate these models into real‐time processes linked to call centers or agencies within Allianz. For example, for a user that is detected to have submitted several quote requests for a car insurance with different parameters online, we can predict the best time to approach her with the right offer that suits better her car insurance needs.
Cross‐ and Up‐selling bundled products to fit the new customer needs. We build predictive models to understand what might be the needs of our customers in order to provide a simplified bundled set of insurance products into one. Our models predict the propensity to buy a new product in the Allianz portfolio (cross‐sell) or to upgrade the current product that the customer owns (up‐sell). This task is easily combined in real‐time with the proactive personalization discussed in the use case above. Indeed, by understanding better the customer needs, we can optimize the best price for upgrades and bundled cross‐selling products such that the customer obtains better coverage for the best suited price. For example, for a customer with a house insurance product that has been navigating on the Allianz website in search of legal insurance, we can automatically personalize content in order to offer a bundle of the two products with best legal service included in the household.
Real‐time monitoring to adapt to current consumer trends. We track customers across different channels and blend data from multiple sources—e. g. from past interactions, transactional systems, Web logs and external data—in order to have single view of what customers are doing and what happens as a result. We create comprehensive views of the digital costumer journey in the form of dynamic clusters and segmentations that can be easily visualized by business experts. Every cluster is explained for the business analyst, e. g. the segment of young family consumer that own a small car and are likely to change to a more spacious car in the near future. Our monitoring application provides dynamic segmentation models and interfaces with Market Management departments, who can then monitor in real‐time trends and address concerns such as conversion of quotes, retention or optimizing parameters of products.
Fraud prevention and fast‐paying of “white” claims. We provide a flexible approach for claim fraud detection, prevention and claim management within Allianz. This is also part of the experience in the experience phase of a customer journey, as customers would like to engage with companies that keep fraud processes under control so to avoid overloads in premiums and get back money paid for claims fast. Internal and external data sources are blended to explore networks of involved parties and compute fraud indicators associated to claims. Predictive models can then predict the propensity of fraudulent behavior in a network and within a set of filed claims. These models are combined with an ensemble of predictions at the level of filed documents, history of claims and geo‐location of incidents. Complementary, with the same data we also build models that decide when claims are not fraudulent and therefore can be paid quickly and fast‐tracked. Fraud analytics is also considered important to improve the user perception, as while on the one side it is important to keep the portfolio clean of fraudsters to protect the candid policy holders, it also improves their user experience as they are paid faster.
In all the above described use cases, a critical issue is to continually do A/B testing to compare alternative versions of the decision process (e. g. a Group A would follow the journey as planned by a predictive model and a Group B would be the control group that would simply follow the regular traditional approach without much influence). Continuously te
sting, evaluation, and adaption in a rigorous scientific way helps to identify what works better and how internal processes need to be adapted to improve even further.
59.3 The Era of New Digital Business Models
The value of chain of digital business models is fully embedded in digital processes. Interface with other processes or entities are digital, e. g., even human interactions occur on smart phones, within apps etc. Further, a digital business model does not copy a legacy model 1:1, but focusses on what is done best in the digital world and may not fully cover or copy the traditional end‐to‐end process. Digital business models are based on auctions, peer‐to‐peer, or crowd‐sourcing, for instance. We use external and internal data to expand the digital journey and pursue innovative business models for our customers and prospect customers. Ultimately, the goal is to rethink the insurance business and identify sources of value in any new way that matters to customers.
Focusing on specific digital journeys for car insurance for example, innovation could be to go beyond motor insurance products fixed to one specific car. Telematics‐based motor insurance products do not quite fit the digital aspiration as they mainly extend the traditional model by adding more fine‐grained data to the underwriting process (driver specific data, such as age, residence, was always included in premium calculation). Further, they lack aspects of the experience phase.
Digital Marketplaces Unleashed Page 90