Can Anyone Perform Advanced Analytics?

A solid grasp of techniques bolsters the important skillset business analysts need for advanced analysis

Research Director, Advanced Analytics, TDWI

I have been thinking a lot recently about the skills needed to perform advanced kinds of analytics—what we at The Data Warehousing Institute (TDWI) often call algorithmic analytics. A popular advanced analytics technique is predictive analytics, although there are a host of others, including link analysis, text analytics, and more. There has been a great deal of discussion in the media over the past few years about democratizing analytics to make it more consumable than traditional analytics paradigms.

Ease of use has been on the minds of vendors as they try to engage increasing numbers of end users to take advantage of advanced analytics. Deployment options for advanced analytics are growing too, such as operationalizing a model as part of a business process that makes the output easy to digest by nontechnical people. For example, a data scientist can build a model that might be operationalized as part of a call center process. The call center agent doesn’t need to understand the model that is working behind the scenes; he or she simply needs to know how to take action on the output from it.

But where is the line drawn when it comes to actually building a model? Can vendors make software so easy to use that a degree or even training in some sort of quantitative discipline isn’t needed? And even if the software is simple enough to produce output, is that ease of use enough?


Top skills for performing predictive analytics

TDWI recently published the best-practices report, “Predictive Analytics for Business Advantage,”* that documents a study in which I gathered information about what the market believes are the skills necessary for predictive analytics. In the study survey, 330 respondents who were both utilizing predictive analytics as well as planning to use it in the near future were asked what skills were required to perform predictive analytics. The following top three skills were cited:

  • Knowledge of the business: 74 percent
  • Critical thinking: 67 percent
  • Knowledge of the source data and how to prepare it for analysis: 67 percent

Of course, these are critical skills for advanced analytics. However, training—in either predictive analytics or training on the software—was surprisingly ranked significantly lower than these top three skills. There is more to each of these skills than meets the eye, which calls for drilling down a bit on several factors.

Knowledge of the business

When it comes to analytics, knowledge of the business is very important because it helps organizations better formulate a business problem worth solving and potentially interpret results. However, that process is more than just knowledge of the business that is required for analytics; it is about being able to frame the business problem in the right way and to keep the global view of the problem in mind.

While some people are experts in understanding their business, properly framing a data analysis problem using analytics tools and techniques is a different skill. This skill requires the ability to articulate the question and understanding the data necessary to answer it, while being mindful of the big picture. And key to analytics is the analysts’ need to be strategic thinkers regarding what problem they are trying to solve and why, what data is needed to solve it, and what they can do with the results.

Critical thinking

Clear, open-minded, and disciplined-yet-creative thinking are the ingredients of critical thinking, which is extremely important for advanced analytics. In analytics, critical thinkers need to be able to rapidly assimilate and synthesize data and information. Additionally, and importantly for advanced analytics, the critical thinker does not take anything at face value—that is, he or she questions the data and the assumptions.

Some people often make the mistake of not questioning data and assumptions. They look at the output of some sort of complimentary analysis provided by analytics software without handling the data, and then don’t think to delve deeper into it. For instance, someone might look at an analysis that says males are twice as likely as females to have a negative sentiment about a specific brand. But they don’t take the time to look at the data that is classified as negative versus positive or even determine how many observations this conclusion was based on. Yes, some software solutions help people handle issues like this, but some don’t.

Academicians argue that college is where people learn critical thinking, and this argument is probably true; some schools even offer courses in it. However, when working with data, being quantitative is important. Analytic thinking with data is a learned skill that needs to be practiced.

An understanding of the source data

Clearly, anyone who is going to build models with data needs to understand the data. Exploring the data is often the first part of this understanding. Awareness of the data is more than simply knowledge of the source data; to do the analysis, analysts need the capability to transform the data in a way that works with the business problem they are trying to solve. If they’re performing some sort of supervised learning in which they have labeled outcome variables, they must think about the data in such a way that corresponds with these variables. They will also probably have to transform the data to either work with the models or to develop derived variables that make sense.


A background for predictive analytics execution

While I have been following the trend toward utilizing business analysts as the developers of models for quite some time, I have to admit that I’m still surprised by the survey results. Is merely possessing the three critical skills cited in the survey to perform advanced analytics enough? While they are important, I believe business analysts must have at least some idea of the techniques they are using for analysis. Why? They are telling the story, and they may have to defend their analysis.

Is a degree in statistics needed? Not necessarily, but I think some training or continuing education in advanced analytics techniques makes sense. In this way, business analysts can get the most value out of their organization’s analytics investment.

Please share any thoughts or questions in the comments.

* “TDWI Best Practices Report: Predictive Analytics for Business Advantage,” by Fern Halper, The Data Warehousing Institute, December 2013.