Big Data Analytics Projects Start with Taking In-House Data Inventory

by   |   March 20, 2013 11:52 am   |   0 Comments

GRAPEVINE, Texas – To get the most out of a big data analytics project, companies start working well before soliciting proofs of concept from technology vendors.

Buying the right technology is important, but companies looking to become data driven first need to consider what critical decisions they make can be enhanced with more information.

A crucial first step: understanding what data you currently have, said Doug Laney, an industry analyst speaking at the Gartner BI and Analytics Summit.

“I am increasingly concerned, or shocked, at all the organizations that have better accounting of their chairs and tables, or other physical assets, than of their data assets,” Laney said. Such an accounting, he said, is about “understanding what information you have, and then what are the possibilities.”

Related Stories

Gartner researchers: predictive analytics to gain traction in business.

Read more»

How UPS trains front-line workers to use predictive analytics.

Read more»

Building an analytical culture for big data.

Read more»

Opinion: why big data projects fail.

Read more»

Laney said this search shouldn’t just include operational data that’s currently used in business intelligence reporting. Companies often have “dark data,” or data that’s being stored and not used, or was only used in a single report. Email archives represent another source that’s already available; analyzing patterns in when emails are sent and responded to can yield insights for sales and marketing, for example.

“The ability to text mine email is something that is really going to explode,” Laney said. “You’re looking for trends and patterns that can be exploited.”

Laney said enterprises can then add other data sources to their inventory. Public data, like social media, weather, census or other government data sets should be considered, as should data for sale.

“There is probably a commercial dataset available in your industry,” Laney said.

Assess How (and When) Data Can Improve Decision-Making
Another important early step is figuring out whether decisions that are crucial to your business can be improved with more data, or more timely data, or simply better data. Bill Hostmann, a Gartner analyst, calls this “constraint-based optimization,” where the focus is on exactly how important decisions are made and what is standing in their way.

Hostmann said that when IT is implementing a data analytics project, “too often they don’t think about the decisions downstream,” so major features of new implementations go unused because they didn’t affect the business.

He cited one bank in Brazil that considered switching to a streaming real-time information system, but realized that most of its important business decisions couldn’t be made until end of business. There was simply no reason to have the analytics information instantly.

Kishore Agasthi, the vice president of information management for global information services and technology at Sodexo, a major food and facilities management firm, recently implemented a new business intelligence and analytics system using Oracle’s Exadata and Exalytics platforms.

Agasthi said mapping out the problems that are slowing or stopping work is the top priority. In his case, long-running batch queries were killing his business analysts’ productivity, taking 40 minutes or longer. Database administrators were swamped trying to make those queries more efficient.

With his new Oracle system, those queries take 30 seconds, freeing both DBA and business analyst to do more work. That simple increase in speed allows for a far better understanding of the massive amount Sodexo’s enterprise data.

“You have to know what problems you’re trying to solve,” Agasthi said. “The understanding of the problem leads to using the right solution. Technology fits at the end – it’s a tool, not an end in itself. I think that’s where IT can get lost.”

Even after he made his technology choice, Agasthi is looking for new problems that better applications or mobile deployments enabled by the new Oracle system. “It’s a never ending story,” he said.

An Argument for In-House Data Skills Training
One thing a company does not need to do is to hire a “data scientist” for the planning and implementation, according to Gartner analyst Svetlana Sicular. Almost all companies have the in-house expertise required to create a data science team.

“Whenever you hear a company say ‘we need to hire a data scientist,’ it’s an indication that they don’t know what to do,” Sicular said.

By creating a team of employees with technical skills, creative minds, business expertise and communication ability, you have access to the same skills that an expensive data scientist would bring to the table.

By not hiring a data scientist, Sicular said, you also give cross-disciplinary training to those team members and begin to grow your own data science experts.

“An initial goal of big data adoption must be learning and growing big data expertise,” Sicular said. “When the technology becomes mature, you’ll be able to handle it with your own maturity.”

Email Staff Writer Ian B. Murphy at Follow him on Twitter .

Home page photo of stockroom worker taking inventory at FEMA facility in Fort Worth by Liz Roll via Wikipedia. 

Tags: , , ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>