Businesses have always wanted a crystal ball to predict the future. Now they can almost have one if they combine Big Data, analytic prowess and deep domain and process knowledge. For the first time, analytics is becoming predictive and can help executives make less risky decisions that affect their corporate futures.
The key is you have to use the data properly. “Big Data by itself does no good. In fact it gives you a false sense of security,” says Larry Freed, author of new book entitled Innovating Analytics. “Without analytics, using Big Data is worse than having no data at all,” says the author, who specializes in retail.
Today, the question on everybody’s lips is: ‘How do we use Big Data to achieve our business outcomes?” says Jon Bischke, CEO and co-founder of Entelo. The Outsourcing Center asked the experts to answer this question.
Outsourcing service providers have been providing analysis since they started collecting data for their buyers. The problem was they used statistics to look at history. Statistics, however, looked at small samples to predict the behavior of big groups, explains John Mattox II, director of research for KnowledgeAdvisors. “The problem was you couldn’t get all the data you needed because you couldn’t reach everybody.”
That is not a problem any longer. Bischke notes that in the last five years “there has been an exponential explosion of data available online.” Adds Mattox, “Predictive analytics has come into prominence because now we have enough data to infer the future.”
Hardware and software were also a drag on Big Data analysis. Mattox says as little as three years ago the hardware and software available could not handle either the volume or the velocity of the data streaming in. Today, both allow analysts “to test hundreds of ideas and look at hundreds of variables at one time. We can create models instead of testing one hypothesis at time,” he explains.
The wondrous possibilities
Predictive analytics can provide valuable insights when used properly. The movie Moneyball dramatically told the story of how predictive analytics revolutionized baseball. Here are a few outsourcing examples.
Recruiting: According to Bischke, the world’s top talent rarely applies for a job. Instead, a recruiter lures the employee away with a better offer. How do you know when someone might be ready to leave?
Entelo created its own search engine that crawls the Web in search of public information on a person’s professional life. Entelo then uses its database to help its clients recruit hard-to-find talent.
The outsourcing provider uses predictive analysis to determine when someone might be open to a new opportunity. It helps its customers hone in on these candidates instead of having to send out shotgun blasts about the opportunity, which could damage their corporate brands.
MoBolt, a provider that utilizes mobile technology in talent acquisition, analyzed applicant use by zip code, says Kshitij Jain, founder and CEO. It discovered in one zip code females looking for jobs in the service sector applied for these jobs from 4-9 am. But in a neighboring zip code during that same time the crest of job applicants were white collar jobs seekers of both sexes. MoBolt helps its buyers with the key insight on these applicants about where and when they are typically looking.
Employee training: KnowledgeAdvisors, a learning and talent analytics company, had a buyer in the health insurance market that was battling an ongoing employee turnover issue;30 percent of its employees left the company each year. The insurance company studied strategies for continuously improving its on-boarding process, new hire satisfaction and associate engagement scores. It then implemented an automated 30-, 60- and 90-day touch-point evaluation process, which identified flight risks for an early intervention process.
The outcome was a dramatic success, with the company reducing its own administration time by more than 70 percent, while also retaining 93 percent of the flight risks after 90 days.
“The employer gained these benefits because it applied a simple continuous improvement process: measure the problem, determine potential causes, apply an intervention and monitor improvement,” says Mattox. “This measurement example is fully applicable to massive data sets and shows similar benefits when companies apply appropriate interventions.”
Selling pizza: Jain met a Google engineer in a hotel bar in San Francisco; they began to share analytics stories. The Google engineer said a pizza company had his new startup analyzing its delivery volume by zip code. The engineer discovered when the weather was bad, deliveries increased threefold. And when there was a game on TV, deliveries increased twofold. So the pizza company created a campaign that only ran when the weather was bad and a game was on TV.
Big Data’s big challenges
John Schwarz, CEO of Visier, says Big Data “is fraught with all kinds of challenges.” Freed adds companies” can’t use the same old business intelligence techniques to evaluate Big Data. You will waste time and money and miss the important insights if you don’t get it right.
Schwarz says the key to getting the right answer is to start with understanding the business user’s questions, not the data. “If you start with the data, you will collect all kinds of things you don’t really need,” he believes. “And if you architect Big Data incorrectly, you make it very tedious to find the answers to the important questions.”
The best way to go about this is to “make sure you have taken the time to define the insights you are trying to obtain,” says the Visier executive. “And only then start to organize your data into a structure capable of providing the right outcome,” he continues.
Schwarz says modern analytics solutions must allow customers to do scenario modeling. This helps them make changes to see alternative futures and then select the one they prefer. The company recently launched a workforce planning cloud solution that allows organizations to leverage data from unlimited sources to predict how workforce requirements (such as future employee costs, geography demographics and consequently recruitment, promotion and turnover patterns) will change.
Mattox of KnowledgeAdvisors says the best way to reduce risk is to ensure the Big Data sample reflects the entire population. “Representative data minimizes risk because you know what’s happening,” he says.
Jain of Mobolt adds “the depth and quality of the data is important.” He suggests going down to two or more levels (the first level being name, gender, zip code and email address) “if you want to assess more meaningful trends.”
Finally, the task is too complicated to do in-house. Next time we will discuss how ConAgra, which has 36,000 employees, uses Big Data to manage its HR and recruiting. It originally did the analysis in house before outsourcing the process to Visier.