Foreword

Customers do not conform to one data type, one channel, or onesize-fits-all styles of marketing. Customers are empowered. Unhappy customers will go to a competitor—and as social influencers, they will take potential customers with them. Organizations cannot afford to be complacent about how well they know their customers. Competitors who are adept at quickly turning data and information into knowledge will have a leg up in the race to attract and retain their customers.

Leading organizations are tapping both traditional and newer “big data” resources to run advanced analytics programs to discover trends, patterns, and other insights. These can fuel new products and services or guide more efficient and effective operations. Although the term big data is often defined inconsistently, it generally refers to new sources of non-transactional, semi-structured, or unstructured data that includes Web clickstreams and logs, machine data, location data, and text. These sources have the potential to deliver what transaction data alone cannot: contextual insights drawn from behavior across channels. These sources help organizations understand and anticipate what behavior leads to a purchase, what types of engagement will keep customers loyal, and how customers influence each other in social networks.

Big data is not exclusive to large organizations. Even small to midsize organizations are confronted with more data volume, variety, and velocity than they can handle with traditional systems. In response, many enterprises today are expanding their customer data architectures to include cloud data services, Hadoop, pre-configured analytic appliances, and data virtualization.

Advanced analytics is essential for deriving maximum value from customer data. Advanced analytics methods and technologies include predictive analytics, statistics, data mining, machine learning, and natural language processing. They enable users to get beyond standard business intelligence (BI) or online analytical processing (OLAP) querying and reporting to explore data for patterns, trends, and correlations. Big data analytics is about applying advanced methods and technologies to derive insights from very large and diverse data sets that often include varied data types and streaming data.

This TDWI Checklist Report discusses steps for realizing value from big data and analytics for better customer intelligence.

Number One: Gain a Comprehensive View of Customers and Prospects Across Multiple Channels

Data is the foundation of any customer intelligence strategy. Today’s empowered customers and prospects generate voluminous and diverse data through activity across channels. These include physical stores, kiosks, field-sales offices, and call centers as well as newer channels such as mobile, e-commerce, and social media. In addition, many organizations want to tap customer or prospect data from activity generated through reseller and indirect channels such as advertising networks. Organizations can gain insight by integrating views of these data sources with demographic information.

Thus, an important first step is to develop a strategy for accessing and integrating customer data. Performing analytics on data drawn from one channel or touchpoint source can be helpful, but you can gain significant insights by analyzing multiple sources. Indeed, a key objective is to gain a single, holistic view of customers and customer segments by integrating data from all relevant sources. This “360- degree” view of a customer’s interactions and experiences across channels allows you to respond intelligently when a customer engages your organization at any touchpoint; in real time, organizations can use advanced analytics to determine the appropriate marketing approach and offer relevant products and services.

Business leadership should identify relevant sources and work with IT to determine which users and functions need to view which data sources. The classic approach is to consolidate customer data into an enterprise data warehouse (EDW) that serves multiple users or functions. Although often the best approach, it is not the only one, nor is it easy; TDWI Research finds that organizations regard EDW consolidation as challenging, not just technically but also for people and budget sponsorship reasons. Fortunately, data warehouse automation and pre-configured analytic platforms are making the technology for consolidation easier.

Options today allow organizations to expand their data architecture to include technology approaches that complement the classic, on-premises EDW but require less data movement and transformation. Choices include data virtualization, which reduces data movement, as well as cloud data services and Hadoop file systems. Such emerging “hybrid” data architectures can support a greater variety of analytics. Organizations should evaluate the spectrum of approaches to determine which ones fit current and future analytic needs.

Number Two: Implement Predictive Analytics to Become More Effective and Proactive

Predictive analytics is an advanced solution that can help organizations anticipate customer and market behavior and respond proactively. Predictive analytics consists of statistical or data mining methods and tools that enable you to develop and score predictive models based on combinations of variables. With BI and OLAP, users are able to slice and dice data to answer questions about what happened or what is happening, but with predictive analytics the focus is on estimating outcomes or “targets” of interest. Data scientists, analysts, and advanced business users can employ methods and tools to test hypotheses or explore trends and patterns in data for answers to “what will happen if” questions, such as factors that, if present, may be driving higher churn. They can also use predictive analytics to “let the data speak”—that is, to explore data to determine the right questions to ask.

Predictive analytics solutions exploit software automation and computing power to derive insights faster. Solutions enable organizations to score models and run algorithms on big data sources and real-time event streams. High-powered algorithms enable organizations to mine data and spot trends, patterns, affinities, and correlations rapidly. In addition, the combination of the solutions’ software automation and big data access allow organizations to evaluate the performance of predictive models sooner. Traditionally, organizations have had to wait much longer to evaluate model performance, which has meant that models become out of date and produce erroneous analysis long before analysts can change them.

Big data is important, but many organizations have had success implementing predictive analytics against less than “big” data sources, including transactions or other business application data. Organizations have been able to operationalize predictive insights so that personnel in contact centers, stores, or heading up e-commerce sites can be proactive in adjusting marketing offers or refining customer-facing processes.

At this step, organizations should define desired business outcomes and evaluate which processes could benefit from predictive analytics to achieve those outcomes. Along with building models, organizations should focus on streamlining data preparation by providing integrated data views and utilizing tools to reduce manual effort.

Number Three: Deploy Advanced Analytics to Enable Personalised Marketing and Engagement

Organizations are often motivated to deploy advanced analytics on customer data because they seek a greater intimacy with customers and prospects. Having grown in size and diversity as a result of acquisitions, business reorganization, or the introduction of new products and services, organizations find that they struggle to retain focus on individual customer preferences, which is critical for ensuring loyalty. The virtues of growth can come at the cost of intimacy. Although it may not be possible to replicate the tight customer connection once enjoyed by mom-and-pop stores where customers were neighbors and known by name and reputation, organizations can improve intimacy through personalization guided by data-driven customer insight and engagement.

Strategic initiatives for personalization—including one-to-one marketing, micro-marketing, finer segmentation, and mass customization—depend on data insights. Predictive and big data analytics enable organizations to explore how customers in defined segments spend over time; organizations can model customers’ propensity to buy additional or more expensive products. They can use analytics to sharpen customer targeting and determine the best time to deliver cross-sell and up-sell offers.

Non-transactional big data sources that capture behavior data from online (and potentially real-time) event streams and clickstreams, page views, and logs—as well as records collected from contact centers and other touchpoints—can be valuable for advanced analytics aimed at increasing personalization. With a 360-degree view of customers’ behavioral and transactional activity across channels, organizations can gain a better understanding of customer preferences as well as the effectiveness of marketing campaigns on customer behavior. They can use this knowledge to create a 360-degree customer experience: that is, an intelligent and unified experience no matter which touchpoint the customer prefers for interaction.

Organizations should determine how they can employ advanced analytics to get beyond simple demographic segmentation and use behavioral insights to improve personalization. Behavior-based segmentation can give organizations a different view of customers based on actions and preferences over time. Operationally, organizations can begin learning how to personalize engagements from the first contact—during “onboarding”—and later through successive engagements. Then, organizations can personalize interactions at each step in the customer life cycle.

Number Four: Leverage Big Data Analytics to Sharpen Social Media Strategies

Social media networks provide exciting channels for marketing and customer engagement. Although Facebook, Twitter, and other major sites have been active for about a decade, most organizations still view these channels as the new frontier for customer and prospect interaction. Thus, they are the land of experimentation; organizations must try different types of analysis and consider the results as part of a learning exercise.

Traditionally, experimentation with a new channel might take years. However, by deploying the software and computing power of advanced analytics on big data, organizations can shorten the cycle. The objective is to listen and learn from social media activity and apply the insights operationally, at the right time, to support marketing and engagement strategies both for the activity overall and in the social media channel itself.

Analytics of social media data can give organizations an external perspective on their brands, operations, and strategies that they do not get from internal sources. One example is the ability to discover and measure influence; leading organizations are analyzing social media data to identify the most influential participants in that channel and how these “influencers” impact their networks of contacts, friends, and followers. By analyzing links between influencers and their networks, organizations can learn the ripple effects of their and their competitors’ messaging. Organizations can also turn findings from tracking and measuring influence into a new and valuable source for determining the value of a customer or prospect. Looking beyond individuals, organizations can refine customer segmentation by examining what topics are prominent in social networks.

Text analytics methods and technologies are the most critical for analyzing social media data. Statistical and linguistic techniques can be employed to interpret sentiments. Using algorithms to filter out noise and factor in common uses of irony, sarcasm, double negatives, and other linguistic elements, text analytics can help organizations more quickly understand the reaction to marketing campaigns, product introductions, events, and competitors’ strategies.

In this step, organizations should determine objectives for taking advantage of the unique qualities of social media data. Keep in mind that most social media interaction is less about making sales and more about building reputation and influencing the influencers.

Number Five: Reduce Latency in Analytics Processes to Improve Real-Time Insights

Speed plus intelligence can add up to a competitive advantage in customer-facing processes. Organizations should be vigilant in identifying and reducing latency at every stage in their flow of customer data and information, from its capture and analysis to presentation and sharing. Reduced latency leads to more satisfied customers; sales and service personnel have current information during engagements and can resolve problems quickly. When sales, service, and contact-center agents have at their fingertips not only routine data about a customer’s interaction history but also actionable insights based on real- or right-time analytics, crosssell and up-sell offers will be more relevant.

Marketing benefits accrue when organizations can analyze data faster and discover insights to users’ applications more frequently. Many organizations seek to get past the limits of seasonal campaigns, which can fall out of sync with what the organization is learning through behavior-based segmentation and social media analytics. Organizations also want to understand marketing attribution in near real time to determine which campaigns are having the most influence in triggering customer purchases.

TDWI Research finds that a priority among many marketing business users is to view real-time data and analytics. To develop solutions, however, data professionals need a clear understanding of what users mean by real time. In some cases, users truly need insights from real-time, streaming analytics: that is, advanced, predictive models and scoring algorithms to observe and interpret patterns in data or event streams as they flow in online behavior, gaming, mobile device use, or sensors embedded in growing “Internet of things” environments. In such cases, analysis of realtime data flows can be superior to that of historical data.

For other objectives, the most important capability is to capture data in real time and then run and refine analytic models and algorithms running against it on a daily or hourly basis. Still other users simply require timely notification of data changes as they are relevant to decisions. Thus, at this step, organizations seeking to reduce or eliminate latency should gain a detailed understanding of user requirements so they can deploy appropriate methods and technology solutions.

Number Six: Improve Data Visualization and Visual Analysis For Nontechnical Users

Advances in data visualization form one of today’s most exciting technology developments. BI and analytics application developers are under pressure to keep up with users’ expectations set by visualization experiences with gaming, mobile, e-commerce, and other consumer interfaces. Dashboards have made data visualization a mainstream element of BI applications. More advanced visualization is needed in the era of big data; as data volume, variety, and velocity increase, critical nuggets of information can go unnoticed in standard tabular representations of data.

Users implementing analytics to explore customer data for trends, patterns, or anomalies need visualizations to gain actionable, operational insight. As expressions of meaning, visualizations are vital for improving how users collaborate on data. Users can share visualizations linked to text annotations as well as other content such as pictures, audio files, and maps to “tell a story” with data through comprehensive, shared views. Visualizations such as callto-action buttons, checklists, or gauges featuring different colors can help frontline users connect insights to actions they might take during customer engagements.

Many analytics tools now offer diverse libraries of visualizations beyond standard line and bar charts. Users can choose from scatterplots, heat maps, tree maps, variance charts, 3-D visualizations, and, for text analytics, visualizations such as word clouds. Some tools allow users to easily drag and drop visualizations into workspaces or dashboards, leaving the data integration “plumbing” to go on automatically, behind the scenes. Many tools enable users to shift perspective by viewing data relationships and comparisons across multiple visualizations. Leading BI, analytics, and data discovery tools enable users to consume big data from within familiar workspaces and dashboards rather than have to drop out into separate tools. Users can also export visualizations from analytics tools into other applications such as sales force management.

In this step, organizations should evaluate tools and platforms to support users’ data visualization needs in their analysis of customer data. Organizations should guide users to avoid cluttering their workspaces with too many visual objects or selecting those that are not appropriate for their needs. Guidance is essential via software wizards or by hands-on visualization experts.

Number Seven: Develop an Architecture That Balances Flexibility With Governance

As organizations become more intelligent about customers’ behavior—and as they increase the quantity of customer data to gain insights—they need to revise their information strategy. Organizations deploying advanced analytics on big data for improving customer insight must also make improvements in data architecture, in particular to more effectively protect and govern customer data.

Ultimately, organizations need an information strategy that balances flexibility and governance. Architectures and policies that are too restrictive will impinge upon marketing functions that must be flexible in how they work with customer data. However, too little structure and oversight could expose the organization to chaos, penalties, and damage to its reputation. Here are three points to consider in developing a balanced strategy:

• Hybrid data architecture. Many organizations seeking to analyze a variety of data types and meet broader user requirements find they need to build out beyond their data warehouse. Hadoop systems and cloud services have potential for adding flexibility. The growing maturity of NoSQL technologies such as graph databases makes it clear that architectures must be open to diversity. Emerging “hybrid” or logical data architectures can give organizations a blueprint for how they can unify diverse technologies. Organizations should evaluate hybrid architectural approaches to determine how they might help serve the variety of analytics requirements.
• Customer data privacy. Data breaches are unfortunately becoming commonplace. As more technology solutions are deployed, it can be difficult to oversee how (or whether) users are protecting sensitive customer data. Organizations need to establish enterprise-level management of customer data policies. Business and IT should collaborate to define customer data privacy policies and monitor how data is used and shared. Managing data privacy by using a well-documented hybrid architecture could be an effective strategy.
• Data governance and center of excellence (CoE). Data governance and/or center of excellence committees can help business and IT leadership work together more effectively. Many organizations are successfully employing such committees to encourage best practices and guide technology adoption. Business and IT work can together within a data governance or CoE committee to direct adoption of hybrid data architectures and oversee customer data privacy.