Sign up for our newsletter and get the latest big data news and analysis.

Lost in Translation: Big Data Innovation is Missing the Point

Business Intelligence (BI) as currently practiced, on premise or in the cloud, with new tools or old, does not work. Why? Recent innovation in Big Data has focused on the wrong problem: building a better mouse trap, rather than innovating the entire approach.

For example, take Tableau or Qlik. Both offer significantly better Data Discovery than the previous generation of tools; however, despite marketing claims around ease of use and a low level of technical expertise needed, these tools are still designed for experienced analysts and data scientists. Unless highly technical, a business user cannot put them to good use.

This problem exists across the entire BI project lifecycle:

First, someone very technical needs to create a data mart (or warehouse or lake). That person is, by definition, a technologist, not a business expert. Therefore, interpretation of requirements is an essential step in this process. Much is lost in translation during this step. The usual answer is to load everything including the kitchen sink into the data mart, which then makes it unwieldy, even more expensive to maintain, and costs a mint to store and to keep current. Creating a data mart takes time, and business requirements change constantly. Therefore, data mart design is usually not current.

Second, extracting insights from the data mart is an iterative process. An insight leads to unanticipated questions, requiring a new insight. Each insight is a result of a number of queries, metrics, visualizations, and stat models. Each insight requires a sophisticated application, one that typically depends on a technical person to create. Since the technical person creating the application is usually also not a business expert, much is lost in translation here as well. And, because many business users lack the analytical or statistical training to define a productive discovery process, the iterations are ad hoc, lead to blind alleys, and lack of understanding of the result.

So what to do?

Like most technologies, BI is a tool on which someone needs to build a solution for practical daily use and consumption by non-technical business users. The future of BI lies in creation of Applied Business Analytics solutions – solutions built to meet specific business needs, to answer specific business questions, which are delivered as a service and at scale to businesses and end users.

For most business processes, there are documented best practices, with differences by industry sector, geography, or size of the organization. These best practices can be interpreted and converted into analytics that measure, forecast, and model future outcomes based on the understanding of the process parameters which influence a given outcome. One can build an application that contains all the metrics, best practices questions, methodologies and models that can answer literally any question in a given user domain (marketing, finance, sales, HR, etc) – putting the answers directly into the hands of the business users and removing the risk of insights getting lost in translation.

However, to successfully meet the needs of its target business users, an Applied Business Analytics solution must overcome several challenges.

First, the solution must provide flexibility for the user to navigate through the questions, metrics, and models, while also providing guidance to ensure that the user follows a productive path. Generic graphs and charts are often not suitable to explain complex concepts. Rather, purpose-designed and interactive visualizations are necessary.

Second – and the greatest challenge – is to underpin the Applied Business Analytics solution with a data structure that is extensible, flexible, and contains verifiably accurate data. A useful Applied Business Analytics solution must first solve the problem of organizing the customer’s data to provide a useful multi-dimensional data schema, populated by current, accurate, and verified data.

Finally, the solution must support the ability for business users to take insights and translate them into action. Most people, even those with college educations, are not trained to use analysis for decision-making purposes. Asking the right questions, having the curiosity to dig into results, having the initiative to address processes that are broken or inefficient, requires skills and motivation that may be missing.

Consequently, even in sophisticated organizations that are data-driven, another ingredient is needed to achieve success with analytics: the organization must enable a culture, top to bottom, for its employees to be inquisitive, to learn the analytic skills, and empower them to make changes in processes that need fixing.

To use a current example, United Airlines’ oversold capacity management,

Best practice suggests that the CEO establishes a target process for improvement. Key success indicators are defined. Analytics are deployed to identify metrics, trends, outliers, and factors that influence the outcomes. Management is required to plan and implement changes indicated by the analytics models that potentially lead to better success. Analytics provide ongoing metrics and trend comparisons. Adjustments are made. Results are posted as a learning opportunity for all.

About the Author

John Schwarz is CEO and co-founder of Visier. John has over 40 years of business and IT experience, and was previously the CEO for Business Objects, now part of SAP. He is the co-founder and CEO of Visier, provider of Workforce Intelligence solutions. The company is based in San Jose and Vancouver.

 

 

Sign up for the free insideBIGDATA newsletter.

 

 

Leave a Comment

*

Resource Links: