Want better analytics? Start asking 'crunchy' questions

A lire sur: http://searchcio.techtarget.com/opinion/Want-better-analytics-Start-asking-crunchy-questions

Nicole Laskowski, Senior News Writer

John Lucker hasn't applied for a new credit card in 20-plus years. But in the last three he has received more than 300 offers from banks, hotel chains, affinity groups -- you name it. It's a tactic sometimes called spray-and-pray marketing, and it's "pretty old science," Lucker, consultant with Deloitte Consulting LLP, said during a Harvard Business Review webinar on leveraging customer data.

If the companies spraying him with offers had analyzed his past and present behavior -- how often he applies for a credit card or how he interacts with their businesses -- they could have predicted his response and saved their prayers for a more likely customer. Lucker's personal experience has convinced him that rudimentary practices like this need to be refined. CIOs, it's time to help your marketing departments implement better analytics.

But achieving better analytics by using predictive analytics is no trivial task, Lucker said. It requires sophisticated tools and techniques as well as innovation and creativity -- the stuff of data scientists. Together, they lead businesses to answer "crunchier" questions -- his label for those "practical and detailed questions about the toughest business issues you might face."

Crunchy questions are aligned with strategic goals, relate to key performance indicators(KPIs), are designed to be actionable as well as informational, and provide foresight rather than hindsight. Examples? Lucker suggested:

• What's the online buzz about your company and how could that impact future sales or products?
• Who are the next 1,000 customers you will lose and why?
• What factors impact customer loyalty the most and why? What can you do to increase loyalty?
• Can you map out the aftershock of a bad customer experience and what could you do about it?

Coming up with a list of questions is just the starting point, though. "You should work to categorize them into relevant groupings and help organize ideas and aggregate them around your overall corporate and customer strategies, as well as desired KPIs," he said.

Smart, simple design

1. Simplify the presentation layer. "Giving someone an overabundance of information doesn't make it easy to consume," Wise said. Customize visualizations based on the audience. Data scientists, for example, have tech know-how and, most likely, have access to different kinds of data than a general user. Also, make the metrics and any action items easy to find.Visualizations can help people grasp complex data, but creating visualizations to deliver that easy-to-digest characteristic can be tricky. In a webinar from The Data Warehousing Institute, Lyndsay Wise, president and founder of the independent analyst firm WiseAnalytics, outlined five best practices that can help.

2. Design for every device. Mobile devices must be factored into self-service business intelligence (BI) apps and visualizations, Wise said. For IT departments, that means following the customer. "Business solutions [should] reflect the type of technology used externally and outside of work," she said.

3. Highlight important data. Figure out why the data needs to be visualized, which can help identify how to represent it. Data points that reveal change over time, for example, are prime candidates for animation, which can help users identify trends and opportunities. 

4. Focus on consistent design. "Develop an internal design style to reflect the branding of an organization and what you're hoping to achieve through analytics," Wise said. It's easier on users who might have to jump from one visualization to another, but it's also easier to maintain and scale out as needed. Plus, you can train the whole organization at once, she said.

5. Don't forget about governance. Data visualizations can help open up data across the enterprise. Incorporating privacy and security features into the design is critical and requires advance planning. "These data factors are important and could affect pricing, cost in development and how data is utilized interdepartmentally," Wise said.

Analytics -- au naturale

QlikTech Inc., a data discovery vendor, is rolling out something called natural analytics. The technology is designed to "use our innate skills and build on what we already know,"Donald Farmer, vice president of product management for QlikTech, said during a recent webinar. Here's how he went on to describe it: Natural analytics is to analytics as using a mouse is to swiping a screen. In other words, users have to learn how to manipulate the mouse, but touch "is a natural experience," Farmer said. QlikTech's "natural analytics" is built to capitalize on pattern recognition, outlier detection, sorting and categorization.