Perhaps the most important aspect of big data mining is determining the appropriate use cases and needs that the platform would address. The success of any big data platform depends largely on finding relevant problems in business units that will deliver measurable value for the department or organization. The hardware and software stack for a solution that collects large volumes of sensor or streaming data will be materially different from one that is used to analyze large volumes of internal data.

The following are some suggested steps that, in my experience, have been found to be particularly effective in building and implementing a corporate big data strategy:

• Who needs big data mining: Determining which business groups will benefit most significantly from a big data mining solution is the first step in this process. This would typically entail groups that are already working with large datasets, are important to the business, and have a direct revenue impact, and optimizing their processes in terms of data access or time to analyze information would have an impact on the daily work processes.
  As an example, in a pharmaceutical organization, this could include Commercial Research, Epidemiology, Health Economics, and Outcomes. At a financial services organization, this could include Algorithmic Trading Desks, Quantitative Research, and even Back Office.

• Determining the use cases: The departments identified in the preceding step might already have a platform that delivers the needs of the group satisfactorily. Prioritizing among multiple use cases and departments (or a collection of them) requires personal familiarity with the work being done by the respective business groups.
  Most organizations follow a hierarchical structure where the interaction among business colleagues is likely to be mainly along rank lines. Determining impactful analytics use cases requires a close collaboration between both the practitioner as well as the stakeholder; namely, both the management who has oversight of a department as well as the staff members who perform the hands-on analysis. The business stakeholder can shed light on which aspects of his or her business will benefit the most from more efficient data mining and analytics environment. The practitioners provide insight on the challenges that exist at the hands-on operational level. Incremental improvements that consolidate both the operational as well as the managerial aspects to determine an optimal outcome are bound to deliver faster and better results.

• Stakeholders' buy-in: The buy-in of the stakeholders—in other words, a consensus among decision-makers and those who can make independent budget decisions—should be established prior to commencing work on the use case(s). In general, multiple buy-ins should be secured for redundancy such that there is a pool of primary and secondary sources that can provide appropriate support and funding for an extension of any early-win into a broader goal. The buy-in process does not have to be deterministic and this may not be possible in most circumstances. Rather, a general agreement on the value that a certain use case will bring is helpful in establishing a baseline that can be leveraged on the successful execution of the use case.
• Early-wins and the effort-to-reward ratio: Once the appropriate use cases have been identified, finding the ones that have an optimal effort-to-reward ratio is critical. A relatively small use case that can be implemented in a short time within a smaller budget to optimize a specific business-critical function helps in showcasing early-wins, thus adding credibility to the big data solution in question. We cannot precisely quantify these intangible properties, but we can hypothesize:

In this case, effort is the time and work required to implement the use case. This includes aspects such as how long it would take to procure the relevant hardware and/or software that is part of the solution, the resources or equivalent man-hours it will take to implement the solution, and the overall operational overhead. An open source tool might have a lower barrier to entry relative to implementing a commercial solution that may involve lengthy procurement and risk analysis by the organization. Similarly, a project that spans across departments and would require time from multiple resources who are already engaged in other projects is likely to have a longer duration than one that can be executed by the staff of a single department. If the net effort is low enough, one can also run more than one exercise in parallel as long as it doesn’t compromise the quality of the projects.

• Leveraging the early-wins: The successful implementation of one or more of the projects in the early-wins phase often lays the groundwork to develop a bigger strategy for the big data analytics platform that goes far beyond the needs of just a single department and has a broader organizational-level impact. As such, the early-win serves as a first, but crucial, step in establishing the value of big data to an audience, who may or may not be skeptical of its viability and relevance.