The star power of analytics has burned even brighter in the last few years, as more success stories emerge.
But while many professionals are happy to see analytics move out of the dusty back rooms to take center stage, it’s also created a new conundrum for those tasked with implementing the programs: It’s still not easy to deliver analytics capabilities, and business expectations for success are at an all-time high. On one hand, the trend has given valuable exposure and funding to some programs. On the other, the data challenges of analytics – accessing, cleansing, integrating, analyzing and acting on results – are arguably greater than ever before, thanks to big data, social media data and sophisticated new analytic requirements. This is uncharted territory for many organizations, demanding new approaches and best practices.
This dilemma inspired industry expert Wayne Eckerson’s new book “Secrets of Analytical Leaders,” which profiles seven leaders from forward-thinking analytic organizations. In it, early adopters share advice from the trenches about running a successful program. The book also inspired informal discussions with other analytics leaders at TDWI’s March 2013 Big Data Analytics Summit in Savannah, Ga. Here are the secrets they shared.
1. Understand the business decisions the new project will support.
Collecting requirements is uniquely difficult in analytics, many said. Many business leaders now know that analytics can help them but aren’t exactly sure how – or perhaps worse, have skewed expectations about what analytics can do for them. This demands a different approach to requirements gathering, explained Stephen Robinson, director of Online Analytics and BI at HomeDepot.com and long-time analytics professional. One technique he’s used throughout his career is focusing on the decisions that the business needs to make that could benefit from analytics. It harkens back to the term applied to this industry nearly 20 years ago -- decision support systems – and still rings very true today, said Robinson. Understanding the business decisions involved can directly connect requirements with a specific business process and measureable business outcome, defying the ambiguity that can plague analytics projects.
“I often ask the business lead to tell me the top six decisions they make that they’d like to support with analytics,” Robinson said. “If they can’t tell me, it’s a clue that we need to go back to the drawing board about the goals of the project.”
2. Evaluate available data, especially unique data, and policies for using it.
Many analytics projects start with the nebulous notion that there simply must be some insight buried in the mountains of data a company interacts with. But before getting too deep into the project, analytics leaders agree that it’s critical to understand exactly what data the organization has, where it is and how it’s structured, with particular emphasis on unique data that’s not available to competitors. Even more important is understanding exactly how that data can be used.
3. Conduct a visual POC for analytics projects.
It turns out that the cliché of a picture being worth a thousand words is true – even more so if that picture is interactive. In her talk at the TDWI Summit, Jennifer Lim, research scientist at Sprint Nextel, explained the importance of conducting proof of concept projects with business users. Keep the scope small, she advised, working with a manageable data set. And whenever possible, deliver visual POCs – either a mockup or, better yet, an interactive version using an off-the-shelf visualization tool. For Lim, being able to build visualizations without having to ask the IT or BI team was key to being able to quickly share and iterate with the business.
“Once you get visualizations in front of them, the light bulb goes on,” Lim said. “The business users will start to churn through the prototypes and can start to apply what you’re sharing with them.”
Seeing the data in action, especially when there is a geospatial or mapping element, can help business users understand what they can get from the data, she said. It may lead to revised requirements, but if that supports a more useful implementation, it’s ultimately worth it. It’s helpful feedback to get early on in a project, when it’s easier to modify the plan.
4. Use analytics tools that support SQL, which your organization probably has the skills for today.
With the explosion of interest in analytics has come new methods and tools, such as R and Hadoop. While new technologies may work well in some parts of an analytics ecosystem, they may present challenges for many analysts and business users who are more familiar with SQL, the lingua franca of analytics. Lim’s advice? Use tools that support SQL to remove any language barriers from analytics adoption. Her organization also leverages systems that support user-defined functions, which can be coded in Java by an expert and then called via SQL by any practitioner.
“Allowing SQL was key, so people could access and use the data. It opened up a whole new realm of users,” Lim said. “I don’t need to know all the Java coding needed to run the analysis.”
5. Holistically consider the present and future business needs, workload and technology options.
Finally, industry luminary and TDWI Summit speaker Colin White advised analytics project leaders to holistically consider the business need, analytics workload and technology options, the key word being “holistic.” Instead of tackling the project in the standard linear fashion – progressing from requirements to project plan to technology selection – White recommends collecting information and considering all of the options up front.
Analytics project leads, in particular, must also stay on top of industry and technology trends to anticipate future needs of the business. With analytics increasingly being a competitive advantage, businesses must move quickly to implement new capabilities and that may mean rethinking the underlying architecture. Analytics workloads are becoming more varied, with different functional requirements across BI tools, ad hoc analytics, big data analytics and data science programs. New analytics projects are increasingly delivered outside of the enterprise data warehouse, leveraging what White calls the “extended data warehouse,” a concept similar to Gartner’s Logical Data Warehouse model. New analytic RDBMSs or technologies such as Hadoop may be a better fit for some projects and may deliver more long-term capabilities to the business. However, asking “Should we get one of these Hadoop thingies?” is not the right place to start, White clarified.
“The tendency is to focus on the technology. That should be secondary,” White said. “We have to look at what the business is doing.”
Many analytics professionals agree that the best advice for starting new projects is to stay flexible throughout the process and avoid the technical tendency to get locked into any one method or solution. You will probably never get all of the requirements in advance of the project. This is true for most technology projects, but especially for analytics, where many are still on a serious learning curve.
Expect change throughout the project. Set a manageable scope and build in “quick wins” that help support real business decisions and give you real feedback about how to improve. Choose technologies that are also flexible and will allow you to refine your program as it grows. And collaborate with others in your organization as they learn more about analytics. Those enhancement requests mean that a system is being used and that people are thinking about how it could be most effective. That’s a good thing! Embracing an agile attitude may be the best “secret” for success in analytics.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access