Building a well-defined use case is the key to successful big data efforts. Without defining the right question, you can’t define the big data resources you need, or build the right analytics to deliver results that provide insight and are actionable. The use case is the centerpiece of big data efforts.
The number one reason cited for the failure of big data efforts is an ill-defined use case. A survey conducted by Infochimps, Inc,. revealed that 81 percent of IT professionals said that big data efforts were their top IT priority, but 58 percent said that previous big data efforts had failed due to “inaccurate scope.” In other words, big data efforts failed because the use case was not properly defined.
With big data it’s always best to look beyond the assumed limitations of the data and ask what you really want to know or think you can learn, and then work to find those results. But to define an effective use case you need to consider factors such as available data streams and desired outcomes.
What We Mean by Big Data Use Case
A use case is typically defined in software engineering as a series of steps between a role or persona, typically called an “actor” and a system to achieve a set goal. In the context of big data assets, a use case is designed to address a specific business challenge by posing a question (the agent) and using patterns or examples (the system) to assess outcomes. Big data use cases are customized to address unique issues and provide actionable answers to business problems.
Some of the most common big data use cases include:
- Big data exploration: Data is not an end in itself. In all big data efforts you want to analyze data streams to extract insight and visualize information about business challenges to facilitate decision-making.
- Customer profiling: Big data can give you a better understanding of your customers, what makes them act and react, how they buy, why they buy, etc. One of the attractions of using big data is you can harness social media sentiments as one of the data streams, since big data can analyze unstructured as well as structured data.
- Security enhancement: Big data can provide real-time data monitoring to lower risk and detect and prevent fraud. Security use cases are continually being extended using social media, email, machine sensors, and other data sources to improve security intelligence and incident prediction.
- Operations analysis: Whether you are analyzing business processes, transactions, customer experience, or machine data from the assembly line, creating use cases to analyze operations can improve operations.
The Elements of a Use Case
- Define the question: Start by defining the business question you want answered. Don’t ask non-specific or open-ended questions; the results will be lots of information that won’t offer a clear conclusion. Instead try to define the business question in terms that will yield actionable insights. For example, ask questions such as “Why are we losing customers?” or “What is the best way to communicate with our customers?” These types of questions invite quantifiable, actionable answers.
- Poll the stakeholders: You want to assemble all the stakeholders to develop the use case, since each has a piece of the big data puzzle. For example, Marketing may be trying to define the question, e.g. “What channel is most effective for selling existing customers?”, but Sales and Customer Support will have insight about potential data sources and IT will have to provide the system.
- Assemble the data sources: The stakeholders will be able to identify the best data sources to help answer the question. Some sources may be stored in the corporate database or somewhere in the enterprise. Other data sources such as historic sales figures may be archived somewhere. Still others such as email threads, social media, or machine data may have to be taken from other sources.
- Build the analytics: Once the data has been assembled, use Hadoop and other tools to develop the analytics to extract the findings.
- Assess the results: The raw information could be interesting, but the insight comes from interpretation of the findings. This is where the business intelligence experts apply their expertise; looking for anomalies and trends in the findings that provide an answer to the question.
- Refine and repeat: Often initial big data efforts don’t yield exactly the desired results. Unnecessary data streams can blur some of the insights, or additional factors need to be added to the analytical criteria. Big data efforts will need to be fine-tuned and repeated for greater value.
So what do you see as the greatest impediment to developing an effective big data use case? Is it defining the question, gaining consensus, finding the right data sources?