Practical Career Advice and Best Practices in Analytics
Being an analyst is not only a technical job it also has a peoples side to it. Given that many MBAs, engineers, and even non-quantitative graduates are interested in Analytics careers, we are sharing some advice on best practices for excelling with Analytics in your career.
This week at the Kellogg School of Management, my MBA students in the Analytical Consulting Lab will complete their projects in which they have examined data for a firm and built recommendations based on their use of Analytics. It is very popular class and a unique opportunity for our MBA students to practice Analytics in a real-world setting. It is also a great class to teach and one that shows how Analytics are being used in healthcare, insurance, sports firms, airlines, retailers, start-ups, and even government. Given that many MBAs, engineers, and even non-quantitative graduates are interested in the career opportunities in Analytics, I thought that I would share some advice on best practices for excelling with Analytics in your career. The advice is applicable to all professionals focused on Analytics.
I find the people side of Analytics is often harder than the technical side of Analytics. I have recommendations on both.
Let’s start with the people side of Analytics:
- Develop Skills in Telling Stories (about Data): This sounds like advice for high school drama class, but the truth of the matter is that by analyzing data you will come up with revelations that need to be communicated to business leaders. The findings will challenge conventional wisdom, defeat internal inefficiencies, and may even draw resistance from colleagues. These communications go much better when the performance of the data is explained by a story. You also appear more objective and judicious, versus dogmatic and tied to a prevailing opinion. Stories can be updated with nuances, too. For instance, a team of my students recently showed that health care customers attrite from preventative medicine more in middle age more than in early or later ages. This was a surprising revelation and one that when related through a story that related the journey of a customer and the interactions with the firm made for a richer discussion with executives and a deeper understanding of what to do to improve the firms performance. People also remember stories better than numbers and will retell them easily more, which will spread your message and fame!
- Explain to Others what Analytics can Accomplish: Often the team asked to do the analytical work is on an island. They alone understand the analytical model. The executives often do not have a deep understanding of the Analytics. This requires extra communication from the people with the analytical skills. Don’t expect your CEO to ask about your predictive analytical model if he or she does not understand it. In such a case, the model will most likely be disregarded. It does not mean you have to present equations, but provide a clear and firm view on “what the model does” or “what happens in the black box,” if you would. I do this with graphics that show Y moves up with X, etc. It is important to be diplomatic and humble in this, too. Don’t question people in explaining Analytics; nobody likes questions that make them feel inadequate. If you explain Analytics well to others, you will soon be the “go to person” for all things analytical and have the trust of your executive team.
- Talk and Communicate with Graphics: We humans remember shapes, colors, and especially faces really well (names are harder – interesting!). In spite of a developed language and script, most people can hardly remember 10 numbers in a sequence. It is one of the reasons we kept phone books and highly value our phone books on smart phones. We are not programmed to remember or even appreciate the differences in numbers. Instead of showing a table, show a bar chart or a line graph. Use a graphic that communicates relative change in the data and allows variance to be seen not calculated! Plus, people love graphics and graphics demand an explanation, showing a graphic means that people will listen to what you have to say. Tables do not demand such attention. It will make your presentations far more compelling and impactful!
- Be More than an Analytics Person: Here is the surprise, if there is one. It is great to have analytical skills and many firms are desperately in need of these skills. However, firms run a business rich in operations, finance, and marketing (and other functions, too). The analytical challenges and opportunities facing a firm are about more than data. These challenges are about how the firm will invest, operate, market, and set strategy. Relate your analytic findings to business implications and strategy recommendations. In this way, you are driving business insights to action. Firms will highly value that and will set you apart as an Analytics leader!
There are some technical things that really great experts with Analytics can do that set them apart. Some of these are based on training and some are based on process. Let’s explore what is needed from the technical side:
- Get Smart about the Techniques and Tools of Analytics: For many people years ago, Analytics was just linear regression in Minitab. The Analytics space now requires a more comprehensive knowledge of non-linear models, data mining techniques, and the tools and even programming that come along with these techniques. Deep Learning, Machine Learning, Clustering, and more – learn how these models work. The good news is that there are many excellent on-line classes for learning the latest analytical techniques and tools. Get up to speed to compete!
- Practice Exploratory Data Analysis: John Tukey coined the phrase Exploratory Data Analysis. In short, this approach is to explore the data and relate its variables BEFORE building statistical models. It differs from Confirmatory Data Analysis in which one seeks to confirm a hypothesis from the data without manipulating the data in ways that might seem to bias the statistical test. Interestingly, rigorous econometricians of Tukey’s time criticized his approach because he introduced (they claimed) bias by just exploring the data. Modern data science owes a great deal to John Tukey. He was absolutely brilliant. His approach has given birth to the rich use of data visualization to search for data trends and anomalies. I think it virtually impossible to understand data before it is plotted and before its natural distributions are understood, graphically. Today, great data visualization is possible with tools like Tableau. Get familiar with how to explore your data and reject analysis that does not show exploration first! Explore your data – FIRST!
- Look for Variance: It seems basic, but it is profound. If you are doing anything in data science or leveraging Analytics, variance is your friend! Without variance or with just of little of it, there is not much useful in a model or your work. Predicting low variance data is not a challenge and will hold few surprising and valuable business insights. So, focus on variance in variables. One of my areas of interest is risk and risk management. It is all about variance, there, too. I use Exploratory Data Analysis to find variance that suggests a need for explanation and variance in variables that can explain business challenges. Remember variance is an absolute measure, but its relative amount is what really matters. Accountants, as I have heard are often disturbed by financial variance, professionals in Analytics should look with glee on variance. It is a gold mind! Make variance detection your fist quest in analyzing data.
- Leverage Percentiles and Rank Order Statistics: In today’s world of Big Data, most real-world data sets are large enough to be used directly without the need for a distributional assumption. Indeed, rank order statistics and percentiles, that is, presenting the kth largest and kth smallest values is an unbiased and uncontaminated way to describe the actual data. Although it lacks the flexibility that comes with statistical distributions, anyone who has watched the Olympics or taken the SAT gets percentiles really well. It is a great way to explain extremes and variances in data. Students often ask for the mean and standard deviation on the class performance of an exam. Why? It makes no sense. It is better to ask for the 50th, 60th, and 75th percentiles, instead. Those will say more about the spread in data at the higher end of the distribution. Relating standard deviations is not natural, but relating data by percentile or rank is easily remembered and quickly leads to an evaluation for assessment.
These important recommendations in being successful with Analytics (and more) are developed in my recent book, From Big Data to Big Profits: Success with Data and Analytics (Oxford University Press, 2015). The book also examines the evolving nature of Big Data and how businesses can leverage it to create new monetization opportunities. Using case studies on Apple, Netflix, Google, LinkedIn, Zillow, Amazon, and other leading-edge users of Big Data, the book also explores how digital platforms, including mobile apps and social networks, are changing customer interactions and expectations, as well as the way Big Data is created and managed by companies. Companies looking to develop a Big Data strategy will find great value in the SIGMA framework, which assesses companies for Big Data readiness and provides direction on the steps necessary to get the most from Big Data.
- Details on First Data Science Job Salary
- How To Become a Data Scientist And Get Hired
- Software development skills for data scientists