IT departments that approach analytics with the same tools and mindset as business intelligence projects often struggle to deliver business value.
Companies have spent billions on business intelligence (BI) software. In 2011 alone, worldwide revenue from sales of BI platforms, corporate performance management suites, and analytic application/performance management software reached $12.2 billion, according to Gartner Inc.¹
Now that more companies are pursuing analytics, many CIOs wonder whether their existing BI capabilities can be applied to these newer analytics initiatives, according to Richard Starnes, a principal with Deloitte Consulting LLP’s Information Management service line. “CIOs want to know if they should invest more in BI, or if they should blow up their BI infrastructures and start over,” he says.
The answer, according to Starnes: Analytics requires different tools, techniques, and organizational structures than BI because it attempts to answer open-ended, forward-looking questions. Too often, Starnes sees floundering analytics initiatives that fail to provide meaningful insights because IT departments are applying traditional reporting tools and development methodologies that just aren’t appropriate for analytics.
Here are some considerations for keeping your analytics efforts on track from the start:
Begin with business goals. Analytics initiatives are more likely to produce useful insights when companies begin with a specific business goal or use case such as, say, finding ways to reduce customer churn by predicting their likelihood of switching to a competitor. While traditional BI dashboards can show companies their churn rates, analytics can explain why customers are leaving and even where they’re going, allowing companies to formulate more targeted retention strategies, according to Starnes.
Establish an analytics office. Because many analytics efforts begin in shadow IT functions, leading companies have begun creating analytics offices, staffed with data strategists and data scientists, to oversee disparate projects, according to Kanishk Priyadarshi, a manager with Deloitte Consulting LLP’s Strategy & Operations service line. Companies typically design an analytics office to drive widespread use of analytics across the enterprise, standardize tools and platforms, and confirm data accuracy and consistency, he says. Priyadarshi adds that Amazon.com and President Obama’s re-election campaign are examples of organizations that have reaped value from their analytics offices.
Don’t hoard data. Priyadarshi says some companies make the mistake of trying to gather all imaginable data to feed an analytics project prior to its start. “Humans are natural hoarders,” he says. “There’s an insatiable and addictive need to hoard data, but it distracts them from answering the questions that truly matter.”
One danger in hoarding data: By the time IT staff have gathered the data they think they’ll need, they’ve run out of budget funding to do anything with it, Priyadarshi points out. A more effective approach, he adds, would be for data strategists to source data according to a specific data acquisition strategy.
Create a data playground for users. Starnes says traditional reporting tools aren’t fully adequate for the world of analytics. They don’t permit the open-ended data manipulation and discovery required to answer forward-looking questions. CIOs should consider seeking out flexible applications like Tableau, QlikView, or SAP HANA Studio to complement BI tools, according to Starnes. He notes that those and other applications built for analytics tend to be easier for business users to work with and allow them to determine cause and effect as well as examine relationships between data sets.
Embrace agile methodologies. Many IT organizations set up BI capabilities for business users following a traditional “waterfall” development approach, according to Starnes. They take users’ requirements, and months later they return with a prototype.
Starnes and Priyadarshi say the waterfall method doesn’t work for analytics. “Because the questions analytics attempts to answer are forward-looking, it’s hard for data scientists to know exactly what data structures to build to answer those questions,” says Starnes.
Moreover, business users tend to encounter difficulty defining their information needs, and the waterfall method relies on business users’ ability to articulate what they want.
“Analytics projects require an agile and iterative design approach with business analysts working closely and collaboratively with end users,” says Starnes. “They need to play with different data sets to zero in on requirements.”
Avoid the trap of perfectionism. Closely linked to the waterfall method, the quest for the perfect data model or algorithm can doom analytics projects. Priyadarshi notes that many IT departments focus on building perfect systems, as opposed to rapidly iterating via agile development methodologies. But the quest for perfection, when applied to analytics, hampers data strategists’ and data scientists’ ability to realize when they’ve made a mistake.
“Analytics is about failing fast and smart so you can recalculate your models quickly,” says Priyadarshi. “Developers usually have to run through several iterations of an algorithm before it’s correct. You simply can’t do that with the classic waterfall lifecycle. Your budget will run out before you deliver a prototype.”
¹ “Market Snapshot: BI, Analytics and Performance Management Software, 2011,” Gartner Inc., May 31, 2012