The Four Fundamental Types of Data Analytics

Engaging with various data-oriented companies that employ data analytics to gain valuable insights and drive their strategic decision-making, it’s essential to consider these four fundamental types of data analytics.

I find it practical to assess which type aligns with your specific data-related task, as misconceptions can be costly to resolve later.

Although I learned about this when I started my data career, and perhaps you did too, it’s easy to mix things up, especially when biased by a preference or enthusiasm for one of these. So, think of it as revisiting it:

Descriptive Analytics:

This is simply about What Happened in the Past? You are trying to focus on driving insights from historical data and past behavior.

I would like to take customer churn as an example, and here descriptive analytics might reveal trends in churn rates over time and identify key factors contributing to customer retention.

Tools typically used include Excel, SQL for data querying, and data visualization tools like Tableau or Power BI.

If there is one misconception to point here is that would be saying descriptive analytics is sufficient on its own for making informed decisions about the future! Remember, it only tells you what happened, not why or what will happen next.

In other words if you think about what the expected deliverable for this descriptive analysis that is often include reports, dashboards, or visualizations summarizing historical data trends and patterns.

Diagnostic Analytics:

This type is trying to answer the WHY question, it aims to uncover the reasons behind past events. So the activity will be digging deeper into the data to understand the root causes of certain outcomes.

In the case of customer churn, diagnostic analytics might discover patterns such as dissatisfaction with a particular product or poor customer service experiences.

You expect to use techniques like regression, correlation analysis, or even root cause analysis. Tools would be programming languages and their relevant packages like Python or R can be utilized for statistical modeling. Now as causality is the difficult part one misconception will be that some may mistakenly assume that diagnostic analytics can always provide definitive answers about causality.

So always remember, correlation does not always imply causation, and identifying true causal relationships often requires further investigation!

Predictive Analytics:

The key word here is future! We would like to forecast outcomes based on historical data. It is like : given this happened in the past, what is more likely to happen next?

In the context of our customer churn example, predictive analytics might predict which customers are most at risk of churning in the near future, and then you can analyze these predictions.

As you might have already guessed, you will use machine learning algorithms and tools like SAS, TensorFlow, LightGBM.

One misconception I often see from business side is that predictive analytics can foresee all possible future outcomes with absolute certainty! The idea is that business side should always be involved in the process so the whole analysis makes sense.

Prescriptive Analytics:

In my honest opinion this type is the most attractive one to the businesses. Simply because it addresses what businesses aim to do with data!

Prescriptive analysis goes beyond prediction to recommend actions that can be taken to influence future outcomes positively!

In our churn example, prescriptive analytics might suggest targeted marketing campaigns or personalized retention offers for at-risk customers.

Tools wise you are expected to use optimization algorithms, simulation modeling, or decision support systems. You might still expect to use machine learning or counterfactuals as a part of causal inference.

One misconception might be that while prescriptive analytics can provide valuable recommendations, it’s essential to consider real-world constraints and uncertainties when implementing suggested actions.

It worth mentioning that human judgment and domain expertise often play a crucial role in interpreting and executing prescriptive insights effectively!