More than a third of respondents to our last study on Big Data utilization (BARC Survey “Big Data Use Cases – getting real on data monetization”) said that finding compelling use cases is one of their major challenges. In this blog I want to present our methodology for identifying and prioritizing use cases for innovation, digitalization and big data.
Use cases: the new mantra
Use cases are gaining more and more importance when it comes to the exploration of ideas, requirements analysis, and prioritization of tasks in analytics and Big Data projects. To identify, structure, and prioritize the most promising cases, we recommend combining two common approaches: conceptual and data-driven.
The conceptual approach uses human intellect and creativity to come up with new ideas. Methods like design thinking help subject matter experts to identify and formulate potentially beneficial use cases. Thinking outside the box supported by creativity techniques has proven helpful in coming up with ideas. The biggest weakness is the lack of validation and substantiation of ideas with data and facts that could lead “theoretically designed” innovations, which are difficult to implement in actual company processes.
The second approach is a data-driven approach to finding use cases. It starts with the data at hand, potentially adding additional external or previously not accessible internal data. Applying techniques like pattern recognition to it, maybe in initiatives like “bring your data” workshops or hackathons, help to understand what is in the data and which use cases can be derived from it. This pragmatic approach works very well for initial analysis of data, but higher-value and successful use cases can be found more easily if a data-driven approach is combined with a conceptual approach. This is precisely the idea behind the BARC “Smart Data Science” (SDS) methodology (Figure 1).
Figure 1: Combining the conceptual with the data-driven approach to identify use cases is a key element in identifying use cases for big data and digitalization.
With the progression of digitalization, companies are increasingly setting up “systems of innovation” and supporting organizations for explorative data analysis. But neither the exact requirements nor the real business value of these new analytic applications and their models can be defined upfront. Instead, many use cases usually have to be examined during the project to see if there is really added value – or nothing at all!
To reduce costs and the failure rate it is therefore helpful to identify and prioritize the most promising use cases early on. In this respect SDS helps in many ways: to select use cases as the foundation of a data or analytics strategy or to steer projects that are initiated to support digitalization and create innovative applications.
How the methodology works
The starting point for the methodology is typically a workshop to collect ideas from domain and data experts from different levels and areas of the business. The goal is to understand what ideas there are for an innovative (or at least different) use of data. At the same time, data sources and business processes are looked at to identify where more, different, or better data could bring value in steering processes, forecasting behavior, or making strategic decisions. Methods like design thinking may help in this situation to extract and formulate ideas, solve conflicts, or clarify ambiguous aspects in terms of available data, tools, skills, or processes.
The next step is to filter and concretize all these ideas into use cases that could have a high business value. All use cases get systematically documented in a use case grid. In our experience it is important to identify use cases that not only create insights but also have an actual impact on the company’s bottom line.
SDS aims to get a description of important aspects of every use case to clarify their readiness. Information gathered contains, for example, the involved data, data sources, required functionality and technology as well as organizational aspects. Yet this may also reveal possible limitations that need to be considered in further activities. At the end of this process, an understanding of the potential benefit and limitations of all use cases is reached so they can be prioritized using a scoring model. The methodology catches use cases that lead to an incremental improvement of processes but also those that have a disruptive impact.
Proof of data value
Highly prioritized use cases in the SDS process are subject to a proof of data value. In this test, domain, analytics, and data experts team up to validate the assumptions and possible outcomes of a use case. It is especially important to achieve rough checks of data and first versions of models quickly. In a “fail-fast” culture, time is of more importance than thoroughness since testing and failing is an integral part of any explorative process. In our experience, once the goal of an analysis is defined and the data prepared, a data scientist can generate initial results within a few days that help to decide whether it is worth spending more time exploring a case or not.
After applying this methodology many times we see excellent results: It helps companies tremendously not only in identifying and prioritizing use cases successfully, but also in establishing an agile approach to validating their usefulness. After all, identifying, prioritizing, and validating use cases is not a one-shot effort but needs to be established as an ongoing iterative process.
For more on how Big Data benefits business, see 1,000 Big Data Use Cases In A Nutshell.