sommai - Fotolia

Analytics team structure can work without data scientists

Despite the value data scientists can bring to a big data analytics team, not every business requires them. Some are using a mix of technology and culture to avoid the need.

When executives at medical benefits management company eviCore Healthcare started thinking about how to take better advantage of real-time and predictive analytics in 2013, they wanted to avoid building an analytics team structure that would stand separate from eviCore's business units.

In particular, Matt Cunningham, an executive vice president who is in charge of operational improvement initiatives, knew from his prior experience as head of IT that building an IT-centric analytics infrastructure can lead to data silos that make it hard to share meaningful information. His goal was to "hand the business back to the business" through the increased analytics efforts. "We should be able to drive the analytics and decision making to the lowest level possible," Cunningham said.

For that reason, eviCore eschewed much of what's often associated with advanced analytics programs. Cunningham hired a team of four analysts, all of whom have master's degrees rather than Ph.Ds. Their primary areas of expertise are in economics and specific business lines like finance. Only one has a background in math.

The company also skipped the market-leading advanced analytics tools, such as software from SAS Institute and IBM's SPSS, instead opting for software from startup Alpine Data Labs. Cunningham said the choice of tools was important in avoiding getting "locked into data science."

Stopping the hunt for analytics unicorns

In reality, there has been little effective integration of good data modeling against complicated data at the business level.
Matt Cunninghamexecutive vice president, eviCore Healthcare

It's a strategy that other companies are also adopting today. As the price for data scientists continues to rise and their ranks remain insufficient to satisfy demand, businesses like eviCore are looking for ways to leverage their data and build out a big data team without shelling out big bucks for what have been described as data scientist "unicorns."

For Cunningham, it also has to do with being realistic about what data can accomplish. He said he's always skeptical when a software vendor tells him about all the great things big data analytics can deliver. There's no shortage of promise when it comes to streaming analytics and unstructured data analysis, but the issue Cunningham raises is how that ties into the business. "In reality, there has been little effective integration of good data modeling against complicated data at the business level," he said.

As a result, Cunningham is more focused on structuring his analytics team to derive tangible value from specific data analysis projects. Currently, he said, the team is working to build better analytical models to predict which medical benefits claims should be paid by insurers and which shouldn't. The goal is to shorten the time it takes to get an answer on coverage when a healthcare provider submits a claim on behalf of a patient. This involves running claims data through a model that looks at factors like the medical diagnosis code, the specialty of the doctor submitting the claim and other clinical information.

Right tools, analytics team structure

For Infectious Media Ltd., a London-based digital advertising firm that focuses on real-time ad bidding, eliminating highly specialized data scientists came down to choosing the right back-end technologies to support a more self-service analytics environment. Daniel de Sybel, the company's chief technology officer, said it used to put all of its data in an Infobright analytical database. But, he added, that had scalability limitations, so a year ago, he and his team started evaluating different options.

They looked at Amazon Redshift but decided that it required too much knowledge around managing resources and knowing how much compute power would be needed for each job. Similarly, Hadoop came with technical challenges that required specialized expertise beyond what Infectious Media possessed, de Sybel said.

The company ended up settling on Google's cloud-based BigQuery analytics platform. De Sybel said it has the scalability of Hadoop with the simplicity of traditional SQL-based relational databases. The development team found they could easily put a front-end analytics system from Looker Data Sciences on top of BigQuery without having to manage the back-end resources so intensively. That has enabled Infectious Media to rely on a two-person team of data analysts to service the company's deeper analytics needs while enabling front-line staff like campaign managers to do more of their own reporting and data exploration. They use the Looker tool to track the performance of specific campaigns and identify potential areas for improvement. It used to take analysts to do this job.

"We've always been data-driven, but the scale of ambition has always been limited," de Sybel said, adding that the new infrastructure enables a broader culture of analytics.

He added that keeping complexity at bay while supporting advanced analytics capabilities was a primary goal of the analytics team structure. Analytics is only powerful when people understand the results, he noted, and if it becomes the sole province of a select few, it isn't going to have the desired business impact.

"The way I view technology is as an enabler," de Sybel said. "It shouldn't be something that confuses people. It should allow your business to do more."

Ed Burns is site editor of SearchBusinessAnalytics. Email him at [email protected] and follow him on Twitter: @EdBurnsTT.

Next Steps

How to create a culture that supports advanced analytics

A data driven-driven culture is needed to derive value from advanced analytics

How to attract analytics talent to your healthcare team

Dig Deeper on Business intelligence team