Advanced analytics projects are inherently complicated. And companies that make wrong turns can greatly escalate the risk of missing out on the expected benefits of an advanced data analysis process – or worse, of dooming an expensive and resource-intensive initiative to outright failure.
Given that many organizations are still maturing in their use of advanced analytics tools and techniques, there’s ample opportunity for missteps. So, where do companies go wrong?
Many problems have to do with not enlisting the right people for an advanced data analytics project and issues related to data quality, according to experienced users and industry analysts. Organizational culture is another big impediment, especially for companies in which end users are accustomed to operating on gut instincts as opposed to a process built around fact-based decision making.
Perhaps the single most prevalent mistake is not fully understanding the nature of advanced analytics and the process involved in formulating predictive models and data mining algorithms, said David Menninger, a research director at consulting firm Ventana Research Inc. “Companies can be overly ambitious or think data mining and predictive analytics can do everything,” Menninger said. “It’s not magic; it’s math.”
Such problems are reflected in survey data, he added. For example, in a survey of business and IT professionals conducted by Ventana last year, a slim 9% of the 308 respondents reported that they were very satisfied with the business intelligence (BI) and analytics efforts in their organizations. Meanwhile, a total of 53% said they were only somewhat confident or not at all confident that the deployed BI and analytics technology was meeting business needs.
Here are more details on some of the ways that an advanced data analysis process can go awry:
Assembling the wrong team. As advanced analytics software gets easier to use and becomes an integrated component of conventional BI suites, it’s tempting to give mainstream business users the opportunity to take advantage of the sophisticated functionality that predictive analytics and data mining tools provide. But that doesn’t mean you can rely on such users to understand the principles required to develop analytical models that will shed light on hidden data patterns and business trends.
Advanced analytics project managers should make sure that their teams include people who have sound knowledge of disciplines such as statistics and qualitative analytics and can properly manipulate the available data, said Rick Sherman, founder of Stow, Mass.-based consulting firm Athena IT Solutions Inc. “Otherwise, it’s garbage in and garbage out,” he warned. “You can get a tool to predict anything you want, but it doesn’t mean it’s valid.”
Invalid or incomplete data. Data quality problems are a major stumbling block for advanced analytics projects, as they are for traditional BI and data warehousing initiatives. As such, organizations need to be sure that data is properly cleansed and normalized, Sherman and other analysts said. In addition, they need to ensure that all of the necessary data, including both historical and real-time information, is accounted for as part of the advanced data analysis process in order to produce reliable and consistent analytical results.
“Companies often underestimate the importance of the availability of data to be analyzed and the quality and trustworthiness of that data, said Dan Vesset, a business analytics analyst at Framingham, Mass.-based IDC. “People coming out of academia are used to having a nice, clean data set to solve problems, but in the real world, that never happens.”
Misalignment with the business. Tucking statisticians and other advanced data analysts off in a room somewhere to crunch numbers on their own without input from the business increases the likelihood that an analytics program will misfire. Mathematical acumen and solid data analysis skills are critical to advanced analytics success, of course. But without a solid understanding of business issues and corporate strategy, it’s impossible to home in on the right variables and correlations for creating accurate predictive and analytical scoring models.
As a result, business, analytics and IT professionals need to work in lockstep on advanced analytics strategies and processes – with many experts recommending that data analytics teams be embedded within specific business units. “When advanced analytics becomes integrated into the organizational culture, and the organization is really using the data to be competitive – that’s only possible when there’s a tight partnership between IT and the business,” said Gartner Inc. analyst Rita Sallam.
Taking a one-time approach. There’s a common misperception that if data analysts plug in the numbers once, they’ll get immediate and reliable results – but typically, that’s far from the case. In most organizations, the advanced data analysis process requires multiple iterations and myriad passes with different data variables to zero in on any kind of actionable findings, analysts said.
And it isn’t just that analytical modeling is an iterative process – an advanced analytics strategy as a whole tends to be a moving target. Just ask Anthony Perez, assistant director of business strategy for the Orlando Magic basketball team. As an organization, the Magic has made great strides over the past 18 months in using advanced analytics software and processes to enable dynamic ticket pricing and some customer segmentation for marketing purposes, Perez said.
But with each new opportunity for applying analytics tools, Perez’s team has had to continuously modify its approach and factor in the need for additional data sources. “It’s been eye-opening for us,” Perez said, “that every time we make a step forward and think we’re ahead of the curve, we realize there’s so much more we can do to get to the place we really envision.”
Beth Stackpole is a freelance writer who has been covering the intersection of technology and business for 25-plus years for a variety of trade and business publications and websites.