Atlantalytics: Demystifying Business Analytics

You may have heard that 68% of statistics are made up on the spot. Or that all models are wrong, but some are useful. Perhaps you’ve witnessed an event that had less than a 1% chance of happening or (ideally) placed a bet with impossible odds and won.

These, among many others, are some of the reasons analytics and statistics can be so terrifying or hard to understand. For example, sometimes a single set of data can be interpreted multiple ways to tell different sides of the same story. Not surprisingly, that makes some people skeptical about how data is used.

One clear example of this is Simpson’s Paradox, which is a phenomenon where a trend occurs in subsections of data but is not present in the entire set (or vice versa). Take COVID data, something we’ve all heard plenty about over the past year. In February, an analysis done by Kugelgen, Gresele, and Scholkopf found that COVID fatality rates were lower in Italy than China for every age group, but higher overall. This seems puzzling on the surface, but occurred because there was a major difference in the demographic breakdown of each country; Italy had a larger percentage of elderly individuals who happen to have a higher fatality rate.

ANALYTICS VS. STATISTICS

In other words, MBAs can’t take data at face value. To improve your confidence in understanding analytics and provide a baseline for your future data work, I’m providing an Analytics 101.

Let’s start with the basics. What is analytics?

Is analytics the same thing as statistics? In short, no. Statistics is the science of collecting and analyzing numerical data to create theory. Analytics takes statistics a step further. In addition to basic collection and analysis, analytics uses the data to discover, interpret, communicate, and predict trends, relationships, and outcomes. Put simply, analytics is more focused on understanding the practical importance of key findings in the data.

Using COVID as an example again, the statistics tell us the number of cases, percent of cases, percentage of fatalities, and plenty of other metrics. Analytics tells us how the number of cases is changing, predicts future behaviors, and allows stakeholders to make informed decisions.

In other words, analytics is essentially statistics with added implications. You may be thinking statistics was hard enough to understand, but don’t stop reading yet. Analytics can be very simple to understand. Just break analytics down into four components, or types:

* Descriptive

* Diagnostic

* Predictive

* Prescriptive

That’s DDPP if you learn by committing acronyms to memory.

1. Descriptive Analytics: What happened?

This is what most people think of when they hear data analysis and should always be the starting point. This consists of analyzing a set of historical data to understand the past behaviors. This typically consists of key metrics that can be tracked over time to show trends in the data and tell a story about what has occurred. For example, most businesses have key performance indicators around metrics like ROI, CAGR, Monthly Sales, and Churn. Think of Descriptive Analytics as looking back on the past.

Danger: Do not jump to conclusions. You can’t make decisions until you know why an event happened.

2. Diagnostics Analytics: Why did it happen?

After discovering a trend or anomaly in the data, you need to know why that is occurring by essentially conducting a root cause analysis to make informed decisions. This can be done through additional analysis, such as drilldowns, crosstabs (pivot tables), regressions, correlations, and many others. For example, it’s great to know monthly sales for an umbrella company was significantly higher in September and October. However, conducting additional analysis to see that these were the months where rainfall was the highest, would provide you with the “why”.

Danger: It is imperative in this step, to ensure that we are not confusing correlation with causation. Simply because two events or metrics are correlated does not mean one caused the other. For example, researchers conducted a study on childhood brain development and found that children who took music lessons has increased memory development. You might be tempted to jump to the conclusion that taking music lessons improves memory; However, if you dig deeper, you will find that music lessons are an indicator of wealth, which usually means better access to resources that help improve memory and cognition. Despite their correlation, the additional analysis casts doubt on the theory that music lessons improve memory.

3. Predictive Analytics: What might happen in the future?

The next step is the largest jump, as we are transitioning from looking backward to projecting forward. Understanding trends and the relationships between different variables allows you to project what the future may look like. Let’s take the same umbrella example. If we can run an analysis to see that, on average, for every 1 inch of monthly rainfall in Atlanta, Umbrella Corp earns $500 in monthly revenue. If we used enough data and developed an accurate model, we could make the statement that 5 inches of rainfall next month would lead to $2,500 in revenue.

Danger: Ensure you are not predicting outside of the range of relevant data, as extrapolating outside of the subset can lead to inaccuracies. For example, if the umbrella data set only had months with 1 to 4 inches of rain, predicting revenue when there is 5 inches of rain could be very inaccurate because the model was not created using any data when there were 5 inches of rain.

4. Prescriptive Analytics: How should we act based on what might happen?

The final step is taking action based on what might happen or what is most likely to happen. This dives into the world of artificial intelligence. A simple way to think of this would be a computer system automatically running a new regression when more data comes available to automatically update a model and then use the corresponding predictive analytics to act towards the best possible result.

Danger: You’ve likely heard the phrase “garbage in, garbage out”. A model is only as good as the data it’s based on.

All the components build upon each other, and you must always go through this series of analyses, starting with descriptive analytics to answer what has happened. After understanding why an event has occurred, you can use your knowledge to make predictions and act accordingly. Now, next time your tasked with using data to gather insights, you can break down your analysis into 4 components and eloquently discuss how you reached your well-informed recommendations. You may also take an extra second next time you see a statistic think about the interpretation of the statistic and what information that may tell you about the future.

Harnessing analytics is now central to how businesses operate as it allows for increased efficiencies, automation, accurate performance measurement, decreased risk, proactive decision making, and new insights. To advance in an organization, MBA students need to continuously make well-informed decisions. By taking statistics one step further and understanding how to interpret and contextualize data into insights, you’ll be able to increase your impact while growing professionally in your organization and career path.

If you want to both understand and use some buzzwords like advanced analytics, automation, machine learning, and artificial intelligence, stay tuned for a future column!

Kegan is an MBA student at Emory University’s Goizueta Business School, where he is concentrating in Marketing, Analytics, and Social Enterprise. Kegan graduated from the University of Georgia in 2016, majoring in Marketing and Statistics. Prior to business school, Kegan was a consultant in the Strategy & Analytics practice at Deloitte, focusing on data analysis and marketing strategies. Outside of work, he is (sadly) a huge Atlanta sports fan and is often running, play tennis, or enjoying time with wife, friends, and family.

DON’T MISS: MEET EMORY GOIZUETA’S MBA CLASS OF 2022