The world of business intelligence (B.I.)
has evolved at a breakneck pace over the past several years, creating a corresponding need for analysts who can access—and make sense of—big and unstructured data. Meanwhile, companies are recognizing the value of that data, which means that datasets (and the need for the infrastructure to store and manage them) are getting bigger than ever. Not long ago, B.I.-related terms such as machine learning or natural language processing were largely restricted to academic circles. Now, those terms appear on job resumes flooding the HR offices of Silicon Valley’s hottest pre-IPOs
. Job titles are also changing, with the “BI analysts” of the recent past now positioning themselves as “analytic engineers” or “data scientists.” At the proverbial end of the day, it all comes down to crunching data for insights about what a particular market will do next
Evolution can be a scary and overwhelming thing. Even so, despite the titanic shifts in business intelligence as an industry, the essential skills required for B.I. analysts remain largely unchanged:
Approach the Data from a Business Perspective
Data and analysis techniques don’t automatically equal good business intelligence. The latter requires approaching analytical problems from a “business” as opposed to a “technical” perspective. Before jumping into nitty-gritty details of an analysis, consider the following: 1. What business questions I am trying to answer?
2. What actions do I expect the organization to take with the key takeaways from the data?
Consider the Audience and End Product
As an end goal, every good business wants to provide better products and services to its customers. Therefore, B.I. insights should allow those businesses to answer questions such as: 1. Who are our existing users? Who could be our new users? 2. What products/services would they would like to have? How could we make their life easier? 3. Are we doing a good job so far to serve them? If not, how can we improve?
Let the Data Tell the Story
A good B.I. analyst lets the data tell the story. For example, in an organization looking to improve user experience with an issue-resolution process, a flowchart detailing the current user experience—with dropout rates analyzed at each step—will help pinpoint areas in need of attention.
Embrace B.I. Tools as an Organization
Thanks to the increased prevalence of B.I. tools, engineers and other IT professionals could find their job descriptions changing rapidly. Some businesses will expect them to perform more like analysts, using data to actively improve the customer-facing elements of a product or service. As part of the rise in proprietary solution packages, open-source software, and online education in B.I. techniques, the technical barriers to becoming a BI analyst have been greatly reduced.
Use the Right Tools for the Right Problems
Even with tools and knowledge in hand, though, many people have a hard time deciding how to best tackle Big Data problems
. “Where do I start?” they might ask themselves. “Which tool do I use in this situation? What data do I collect?” This is where that aforementioned business perspective becomes critical: always start with the business problems you are trying to solve. Another key question to ask: How accurate do I need to be? Some business questions, including classification problems like optimizing garbage collection, do not require extremely precise data. On the other end of the spectrum, drug makers need to be incredibly accurate in dosage manufacturing, or disastrous results could occur. When examining problems that require extreme accuracy, you need to decide which machine learning algorithms (such as logistic regression, neural networks, or support vector machines (SVM)) will suit the task. With large amounts of correctly preprocessed data, all these algorithms tend to provide similar results. For problems that don't require a high level of precision, simpler and widely available algorithms (i.e., logistic regression) often end up used. Be aware that no B.I. tool will provide accurate results without relevant and correct data. Strong data preprocessing skills in SQL
or HQL are a plus. Don't try to reinvent the wheel. Very complex BI solutions can be found on the Internet, varying from open source solutions and code libraries to toolkits. You can also search within your organization for B.I. apps you can use. Oftentimes larger organizations possess a large variety of enterprise tools, but few people know about them. Conclusion
There has never been a better time for B.I. analysts (and analysts-to-be) to learn and apply new tools and skills. But with so many tools and data in play, it’s easy to lose sight of business intelligence’s fundamental purpose: to allow companies to serve customers more effectively. When companies do that, they succeed—and their employees have more fancy B.I. terms to put on their resumes.