- Raw Data, the New Oil
- Top 10 Business Questions for Analytics
- Vital Lessons Learned
- Linking Analytics to Business
- Conclusion
This chapter is from the book
This chapter is from the book
This chapter is from the book
Linking Analytics to Business
Instead of viewing analytics as a tool or a method, this book focuses on analytics as a process.
Business Analytics Value Chain
Rather than focusing on a single stage of the analytics value chain shown in Figure 1-2, this book takes a holistic view on the entire value chain (or value “ring” in this case) for continuous value creation. The focus is on the final business outcomes and their continued improvements using scientific test-and-learn methodology. To achieve sustainable wins over time, it is important that the process be run in a continuous fashion (this is the BAP that will be described in more detail in Chapter 5). Here’s a brief description of the five major components:
)
Figure 1-2 Analytics business value chain
- Start with business questions—All BA processes must start with a valid and high-value business question or idea. Even when the team is making a decision regarding IT, data, analytics models, and executions, it should not be made in isolation and apart from the business considerations. Even though the original business premise might sometimes need to be modified, it should always focus on business outcomes.
- Conduct data audit and augmentation—While the business objective is being set, a quick data audit can help determine whether the business has the right data to accomplish the objective. If not, additional data has to be acquired or derived to augment the existing data. Sometimes the costs of data augmentation can be too prohibitive in the beginning, and a scaled-back business objective can be created first to validate the idea and value before undertaking a full-scale project.
- Extract knowledge—The main goal of the analytics exercises, including simple business intelligence (BI) analysis and advanced analytics modeling, is to extract useful business insights, patterns, and knowledge from the data. This is usually an iterative process. By answering the initial question, a good analytics team often generates several more questions. For example, when you see a group of customers exhibit a certain behavior, your next questions may be these: Who are they? Are they high-value customers? How long has this been the case? Is this a recent phenomenon? By peeling the onion and answering the sequence of questions, you can reach the core of the problem and uncover something that may thus far be hidden and unknown.
- Test insights and hypotheses and knowledge management—Once an insight on the predictors’ impacts on the business outcomes is obtained, it needs to be tested for causal effects. To do this, the scientific testing protocol known as the design of experiments (DOE) should be used with suitable control groups. No insights should be taken at face value, no matter how intuitive or elegant they are. Once validated, the insights and effects must be saved in a knowledge management system for sharing and reuse.
- Execution and optimization—Analytics insights must also take into account how they are to be implemented. The lever settings and their respective effects can be used to help optimize the analytics efforts. The validated insights and optimized lever settings can then be used to permit the business to ask the next level of questions. This begins another cycle of BAP.
Integrated Approach
Instead of viewing the various aspects of analytics techniques, business problems, and Big Data applications in isolation, I approach this by bringing everything together in the right sequence to produce the desired outcomes. This is why the book is not organized by analytics techniques, but instead by the way business problems are solved through analytics.
Hands-on Exercises
By embedding the entire analytics process with real business data within a powerful and modern advanced analytics platform (KNIME),7 motivated business readers can then try real analytics on their own data to answer their question.
Reasons for Using KNIME Workflows
Based on the following reasons, I use KNIME workflows for all analytics examples and exercises in this book:
- Advanced, powerful, and free—As will be shown later, KNIME is not only powerful but (because it was developed recently) also incorporates the latest advanced analytics models, IT innovations, and user-friendly interfaces (for true click-and-drop analytics model development). Best of all, it is in the public domain and free (except for the server version).
- Holistic and integrated views—Some readers might want to see how the results are produced within an entire customer relationship management (CRM) or marketing or sales analytics process; others might want to drill down to a particular stage of the workflow to see how the data is transformed and/or model parameters are set and results produced. More adventurous readers might even want to vary the parameters and see how the changes affect the outcomes. The entire workflow can be viewed at the various levels of aggregation.
- A single unifying enterprise workflow across silos—In an ideal case, at the top level, it is strictly an enterprise business workflow. Below it lie the various sub-workflows that can correspond to the analytics workflows relating to IT, finance, marketing, stores, call centers, logistics, supply chain management, and so on. Drilling further down, you can then isolate the various components of the BA process. By embedding the actual analytics and reports within the metanodes of the sub-workflows, various stakeholders can then focus on the particular level and workflows they are interested in and be able to literally trace any questions to their sub-workflows to gain a holistic view of the big picture.