Tapan Patel — SAS Institute
Tapan Patel is Principal Product Marketing Manager at SAS. With more than 17 years in the enterprise software market, he currently leads global marketing efforts at SAS for Business Intelligence and Predictive Analytics topics and SAS Visual Analytics and SAS Visual Statistics products. He works closely with customers, analysts, press and media, and thought leaders.
Solver International: What functions in a company are needed to support or develop business intelligence?
Tapan Patel: Typically, business intelligence involves a variety of functional areas. First and foremost, when it comes to establishing the objective, or defining the problem and how you’re going to measure whether you’re successful or not, you need to have the sponsorship, the support, and the guidance from the business unit that is interested in using the business intelligence technology. Secondly, traditionally, IT has been involved in creating, supporting, and maintaining the business intelligence assets, content, and data which are needed to create either the reports, dashboards, models or explorations.
Now, over the course of the last four or five years, the role of IT in that context has shifted towards enablement. They are there to make sure that governance is maintained, make sure that the proper data are provided for the users who are building the content. Rather than a producer, they have become more an enabler.
And then, finally, there are the users or content creators. I like to say that all consumers want to become producers. All of them are creating, either building interactive reports or updating our dashboards, or creating visualizations or explorations to find important relationships. For certain a class of users, like data scientists, for example, or data mining specialists, they are building predictive, prescriptive and diagnostic models.
I see three key roles that are important for any business intelligence or analytics initiative. Organizations need business executives and their management to champion and sponsor the project. The role of IT has been more in terms of ensuring governance, making sure that they have the right scalable architecture to serve the needs of a variety of different users and to grow systems in an incremental fashion. And then there are the users, who are all producing and sharing content or insights, depending on the types of problems they’re trying to solve.
SI: With the development of dashboards and the simplification of the user environment, it seems that people don’t even care what’s going on behind the data. They can still develop an ability to do analytics.
Patel: When organizations need to create predictive or prescriptive models, they are doing so to solve specific sets to problems (e.g., risks), finding new opportunities, or optimizing costs. Not all business analysts, and certainly not all business users, are going to have the skill sets to build and interpret the analytical models. There’s a certain class of user who is going to use analytical tools, techniques and methodologies to build and deploy those models. Given that those skill sets are scarce, more and more analytics software is becoming automated, smarter and easier to use. The goal is: “How can I reduce the time needed to build the analytical models, and quickly go through the entire analytic lifecycle? What are some of the steps that could be automated as part of the analytical life cycle?”
For example, can I make use of (self-service) data preparation tools to join my tables or suggest hierarchies? Can I automate some of the feature engineering aspects of the model building? Or can I smartly identify which are the most important drivers for building my analytical model? These will make life easier and more productive for data scientists or information analysts who are hard to find, have enormous workloads, and are expensive, in building analytical models.
SI: Business intelligence is a growing investment for large companies, and smaller companies are starting to look at it as well. How can a small company benefit from developing BI?
Patel: Irrespective of whether it’s a large organization, a small, or a mid-sized organization, our businesses, our public institutions, and our economy, everybody is influenced and affected by the data we are collecting and everybody is interested in getting value out of using BI and analytics. When it comes to small and mid-sized organizations, they have tight budgets, a shortage of IT skills and talent, and often rank low on the scale of analytics maturity. They may not have enough people to prepare, integrate and clean the data—which is absolutely critical for BI and analytics initiatives. It is also critical to identify high priority objectives, business problems or customer issues to achieve and solve with your business intelligence or analytic strategy.
Next, for each business objective, identify metrics that are specific and measurable. Monitor the performance and measure the business outcomes from their BI and analytics initiative. Start small, but even with a small project, measure it, continuously measure it, to see whether it is achieving the goals and expectations from your BI and analytics program.
Identify where you are and what is your current level of analytics maturity. Are those skill sets available internally within the organization? Do you need to develop those skill sets or train people, or can you outsource it or can you bring in external talent? Can you bring in help from local universities to fill in those critical gaps and achieve a level of maturity such that it will help you move from Point A to Point B?
And finally, since they don’t have big or mature IT organizations, they should look into investing in cloud-based BI or analytics programs. With cloud-based BI, more users in small and mid-sized organizations can quickly start creating content and reduce time-to-insights. IT can focus on content governance, privacy and data preparation tasks associated with the cloud rather than selecting and maintaining hardware.
SI: We seem to have taken the approach that we’ve got to teach the ability to use the tools as opposed to developing the tools. That is, the MBA program isn’t an IT program. They want to have the capability to understand how to use the tools that are available, and some tools are very difficult. How does SAS provide a program approach for MBA students that will make it easy for them to learn how to create and use business intelligence?
Patel: In any MBA program, especially one focused on a career in analytics and BI, we have a few programs successfully using SAS products in their courses. Focusing on the right type of user, and providing the right tool, is key to quick adoption and success. SAS Visual Analytics gives the student an ability to do BI and analytics in a self-service and interactive manner. They can prepare data that are needed for creating interactive dashboards or reports, and identify why something has happened. They can intuitively create explorations to find outliers, or find which segments of a customer population they need to target, or they can find the relationships between different variables that are important for future outcomes.
For some students who are interested in predictive analytics but who do not want to build something from scratch, we have the capability to provide a basic set of self-service analytics built into the product. For example, if they want to do forecasting, or if they want to build word clouds to do text analytics, or if they want to do scenario analysis, these features are available to them. The software picks the analytical algorithms for you and gives you the results and explains those results to you in a quick, easy, and simple fashion. That’s an example of how universities, offering MBA programs, are using SAS Visual Analytics for their BI and analytics curriculum.
SI: When you were in your MBA program at North Carolina State, did you have this type of software and capabilities available to you, or was it still pencil and paper?
Patel: That’s an interesting question that brings me back to the good days. We did have software to use! At NC State we were using JMP software, which was a data visualization and statistics software. It was a desktop-based tool and we were learning different statistics-based visualization techniques for data analysis. We were also using different statistical techniques to identify what was important, what were the key drivers, and predicting future outcomes.
SI: Let’s talk about business analysts in a company, looking at the ways of developing business intelligence; where are they going? What’s going to happen in the next, say, five years so they can plan and be ready for the next change in the way BI is applied?
Patel: Based on discussions with customers, what many organizations are looking forward to is streamlining the data discovery and deployment aspects of BI and analytics. They want more efficient and effective ways to get value out of different data and analytics initiatives. Besides technology, they want to understand the role people and processes play in success from data and analytics initiatives.
What we are seeing from a data standpoint is the rise of self-service data preparation. This is the concept where more and more companies are looking to put data preparation aspects in the hands of analysts and business users, rather than limiting it to just IT. Identifying how quickly and easily they can access data, prepare data, and transform it so the data is readily available for downstream BI and analytic needs.
IT does not need to be sidelined. In fact, business and IT will have to work together to understand how self-service data preparation capabilities can affect each other’s roles and responsibilities. IT is definitely getting involved in the hardcore data cleansing and data management tasks, but more often, the business analysts, data analysts, or data engineer roles are getting elevated and staffed to do data preparation tasks in a self-service manner. They can access, blend and prepare data quickly to shrink the time needed for downstream BI and analytics content creation and sharing.
When it comes to the data discovery part of the life cycle, how will the next generation of capability available to business users, analysts, and data scientists make them more productive and effective? The software can automatically find and visualize relevant findings, correlations or exceptions, segments or predictions. And once they automatically find it, with help from natural language generation, how can you explain the insights to the business users and the analysts to assist in making the best decisions?
And, finally, when it comes to the deployment aspect, I think more and more companies are asking the questions, “We have found these insights. We have developed the best champion model. How can I quickly put it in the hands of more people such that they can take advantage of it? How can business users, the sales teams, and the frontline people make use of these insights to make better decisions? How can I integrate the modeling results into my operational systems? And once I get new data or as my business scenarios change, how can the data discovery deployment cycle be quickly turned around so I’m not missing opportunities for revenue generation or getting new customers?
SI: You mentioned cloud computing as a benefit for a company that didn’t have or didn’t want to develop a large IT infrastructure. What other technologies are coming that are going to make a change in the way BI is done?
Patel: The interest and adoption of cloud computing for BI and analytics is growing fast. It is suitable for organizations that are interested in quickly bringing results to the table when it comes to BI and analytics. It is becoming important for companies who have their data readily available in the cloud. It is also important, as I mentioned, for small and midsized companies that don’t have the full-blown IT or support staff that is needed for traditional, on-premises deployment.
Even so, the infrastructure part of BI and analytics is an important aspect. Without the right infrastructure or architecture, you are not going to grow from 10 users to 100 to 1,000. Today you might be doing structured data; tomorrow you might want to combine structured and unstructured data. Today, you might only have simple workloads, involving basic reporting and dashboards, but in the near future you would like to handle complex and diverse workloads involving data discovery and analytics. Having the right infrastructure and architecture is equally important as you grow your footprint when it comes to BI and analytics.
The role of IT should not be minimized. Cloud computing, for example, brings up concerns of privacy, regulation, data preparation. For example, accessing and integrating on-premises data and cloud data is a fundamental question for organizations and IT to solve as they embark on their journey to cloud computing. The cloud is suitable and relevant for many customer-facing BI and analytics applications, and hence the interest in cloud computing is driven more by business units. Business also has the upper hand to start their new projects or move their existing one to the cloud because they have the budget and power to do so.
Another trend of interest to IT is embedded BI/analytics. Once the insights are created, they need to be shared through different channels, like mobile, Microsoft Office, portals, etc. in a repeatable manner to ensure consistency and governance. When those insights are made available to the frontline staff, sales teams, or executives, especially if they’re working online, their first order of business is, “How am I going to collaborate? How am I going to interact with those insights? If I have concerns, can I ask questions right inside the Microsoft Office application or the mobile application? If I want to ask a question through a comment, can other users also look at that comment and the snapshots of the insight associated with that comment? And can I quickly interact with other staff such that all of us come together and collectively take a decision?”
Another area where we are seeing interest from IT is around visualization and querying data from streaming sources (sensors, log files, machine data, social streams, weather data, etc.) especially with the growth of the Internet of Things (IoT). Suddenly we are talking about analyzing and visualizing large scale data volumes and complex data in realtime. Viewing data flows in realtime and understanding which signals are important or not within individual data streams is valuable. It must help the operational systems to process anomalies and proactively take corrective actions. Some use cases might be interested in dashboards that update themselves as new data comes in and sends alerts for abnormal readings.
SI: What would you recommend to people who are, let’s say, high school students or undergraduate students, looking to pick a program to go into the business intelligence and analytics area? What criteria do you feel students should look for in a university or college program?
Patel: Students who are interested in a master’s program in analytics, for example, will find different universities have varied expertise, focus areas, curriculum, resources, and depth and breadth of their faculty. However, I would also suggest students consider how the university collaborates with the local business community to bring them real-world problems to solve. Now, I understand that those opportunities will be much more common in the second year of studies, but are universities participating with diverse sets of companies that can help you understand the context of the problems they’re bringing to the table or the specialty you want your future career to focus on?
Second, business intelligence and analytics tools, are a commodity nowadays. You can either take the path to learn open source software, commercial software or both. Hence finding a program that offers you the flexibility to choose is important. You also need a faculty and a program that explains how the data is structured and prepared, and how you can understand the data relationships. Data management is an important topic to look at for any analytics initiative and is more than half the battle students will fight in their analytics career. Look at whether the program gives you exposure to different data management principles.
Third, the program needs to focus on teaching how to frame the objectives and the problems that need to be solved. How you can identify the problem in many different dimensions is also a valuable skill when it comes to analytics. It’s not just about analytical techniques and methods, but which ones to use to solve the problem and that requires an understanding of the problem and the objectives. An understanding of how to measure success is equally important – in terms of specific quantitative and qualitative goals to prove whether you have achieved the laid objective or not.
Fourth, after you have built your best analytical model and come up with relevant insights, you will need to learn the communication skills to explain the results to the business teams and point out the predicted outcomes once the model is put into production. Learning the soft skills, in addition to your passion, creativity, and curiosity, will put you over the top among your classroom peers now and potentially your work peers in the future. Select the program that gives you the chance to bring these experiences, either through projects, through real-world experiments, or curriculum.
And some of the other things which are also important: does the program take you outside of the classroom and allow you to participate in hackathons or in the Data for Good initiatives within the local community? The latter programs will help bring you a perspective and experience that you will not get in the classroom.
SI: BI is a big, growing industry so it’s important to make the right choices early on and not wait until you’re in your MBA program to find out you made the wrong choice.
Patel: Yes, that’s true. That is very true. Because the industry is moving fast, there are new perspectives being brought in. There are new ways to use data and the insights you’re getting when you put it in the hands of different types of users with different types of problems, whether you are incrementally improving an existing problem or whether you’re trying to find or chase a new problem. It’s a challenge, no question.