October 15, 2015
World of analytics include data collection, to modeling, to artificial intelligence. Knowledge of business analyst would also move from set of skills to another during course of his/her career. Different tools provide expertise to solve different kinds of problems, and different companies focus in working with different domains and analytic functionalities. Putting all this together, one may view analytic abilities spanning a spectrum. So, what is that spectrum, how does it matter?
Knowing current region on analytic spectrum can come handy for analytic professional and analytic firms, both. While firms should aim to cover whole spectrum in their offerings, analyst should aim to move along the spectrum to be master of analytics. This post will introduce you to the spectrum of analytic capabilities, so that you know where you are, and where you want to be.
When you initiate a business analytics project, you often do in context of solving a business problem. Unlike in academics, professional data scientists often have goal in mind which will eventually add to bottom line. Solution to that problem is often attempted as answer to certain relevant business questions. And while for each problem and project, many questions can be asked, questions themselves fall into different realm of analytic offering region.
Questions which describe the data, often through summarizing and aggregating data by various cuts, constitute Descriptive Analytics. Goal is to understand what data is saying about prior known dimensions and task involves counting and other metrics in different form (for example: pivot tables). This is often starting point of business analytics and attempts to make sense of all data gathered. In most businesses, this task constitutes largest chunk of analytics, though human efforts spent may or may not be large since such tasks are often automated.
Questions which tries to understand why something happened or is observed in the data, form next level of Diagnostic Analytics. Goal is to find out reasons for observed data and tasks involve hypothesis testing of various potential reasons, finding right dimensions for aggregation and splitting the data, and looking at patterns in the data. Business understanding and basic statistical knowledge become crucial for solving these kinds of problems. Most analytics jobs lie largely in this region of spectrum.
Questions which attempt to forecast or predict fall in the domain of Predictive Analytics. What is predicted is supplied by analyst, and data is mined for patterns to model the future based on past. Many professional analytics firms operate in this part of spectrum. Goal is to forecast future outcomes with various degree of confidence under various what-if scenarios. Solid understanding of machine learning methods, modeling assumptions and best practices, statistics, and tools beyond Excel such as SAS, R, SPSS, Python are almost always necessary.
While predictive analytics can provide glimpse to future under different actions, they do not advise on actions themselves. Prescriptive Analytics goes beyond prediction and recommends best set of actions for multiple entities looking holistically on all constraints, business requirements and goals. At this region of analytical capability, knowledge of optimization and decision making algorithms/tools become crucial. Only very niche organizations and business can provide and consume prescriptive analytics.
Last and holy grail of analytics is called Pre-emptive Analytics. Unlike prediction and prescription analytics, which try to solve the problem post-facto, Pre-emptive Analytics watches over all areas of business and customers and constantly anticipates and solves a problem before problem even becomes apparent. Very few organizations can really claim to operate in this span since it requires completely integrated data, feedback loop, and Artificial Intelligence built into whole system with limited human intervention.
Apart from advancement in analytic capabilities reflected in analytic spectrum, other orthogonal dimension which affects your skillset is: Who are your clients? Often, analytic companies can be classified into third-party analytics company – who provide services to other companies – and captive analytics company – who provide services to other departments within own company. Former often has more variety in work, though may still have team specializations. Later may provide more opportunity for domain expertise.
In other way you clients will impact your analytic capabilities is by posing the right set of questions. Some clients, mostly new to analytics, feel uneasy trusting a complex “black-box” model to make decisions, while others, mostly those who have benefited from analytics in past, are more open to new and possibly counter intuitive ideas.
Lastly, some teams focus on providing similar analytic solution to different clients again and again and other teams focus on providing different types of solutions.
First types of teams go really in depth into problem solving, often have detailed processes and checklist for taking on projects, invest heavily in advance analytics, and usually work with customized tools and partly or fully automated analytic development. Analyst working on these kinds of projects can expect to become master of that domain. This may however be accompanied by bit of monotony, though in practice each project is different and astute analyst will find opportunity to learn.
Second types of teams have more flexibility and variety in work, which alleviates boredom but introduce challenges of solving different problem, navigating different data structure, more custom work, and data exploration. Often analyst working in these teams will have wider exposure to different business domains and sub-domains but depth and business knowledge may be limited.
As world moves towards Big Data, Artificial Intelligence and Internet of Things, need for seasoned analytic professional working at advance level of analytics spectrum remains highest in history.