Types of Data Analytics
A Short Thought About Data
At first, data is some unconnected information about anything that we can find anywhere as long as we are interested in what it can tell us. I promise it makes sense if you read it once more.
For data to be found, it needs to be seen and registered. Regardless, it is there.
We usually think about data as isolated snippets of information. And it actually is. For instance, the coordinates of a point in the globe, your name and interests, how many times you drink water per day, how much you enjoyed a certain movie or book, how the temperature changed yesterday, how many times you buy coffee per week (and from where), what type of music you listen to the most... they are all data. But they would not be found if no one was never interested in knowing them.
With a lot of those snippets about you, for example, we can even identify who you are, what you like or what you WOULD like (a little of Black Mirror in this one). This is actually what has been being done by the huge companies for a considerable amount of time (I am pretty sure Amazon has reached to you suggesting some products for you to buy. Some of them you did not notice you needed by the way). Now, more and more companies are relying on data to develop their business.
Even you rely on data to make some of your decisions. You go to that restaurant because you know the average rating of the place is really good. You set aside 2 hours of your time to write because you know on average this is the amount of time you need (based on previous experiences). You study for your college exams because you know you will have a better chance to pass it if you do(this is fair correlation and I hope you believe it). In fact, the more you do something(or go through something), the more you get better in handling it. And this is your brain collecting data and improving your decisions 🤓
With this being said, we can imagine how powerful data can be. Identifying patterns and explaining phenomenons are some of the most valuable things data can do for us. But for data to be powerful, it has to be properly used, analyzed and connected. Or they will be just ... lost information snippets with nothing to tell us. Or worse, it can be connected in the wrong way.
Civil Engineering doctoral candidates have been eating a lot of mozzarella cheese, huh? Source: Spurious Correlations
Data needs to be translated into meaningful and useful information.
Types of Data Analysis
This understanding of data and analytics usually comes in four key categories that are built upon each other and are linked to a respective question.
- Descriptive Analytics
- Diagnostic Analytics
- Predictive Analytics
- Prescriptive Analytics
As data analysis develops and gets more complex, it gives back more value.
Types of data analytics. Adapted from Principa
Let's explore each type, shall we?
The main questions here are what happened? and what is happening?
Descriptive analytics is the foundation of data insights and from where data analysis starts. It summarizes past and current data, describing the current state of what we are interested in. In business applications, it usually reflects in building dashboards that include KPIs, revenue reports and sales overview.
The main question here is Why is this happening?
Diagnostic analytics focuses on the underlying reasons for the outcomes from the descriptive analytics. It also looks upon past and current data.
With the "whys" answered, better decisions can be made (and understood), both for changing the current state or for maintaining the way it is.
The main question here is What will/might happen?
Predictive analytics tries to describe the future by using the past/present. It is all about forecasting through predictive models that correlate causes to actions/states.
For instance: which leads have the best chance of converting for our current business, how much we are going to sell this month, what are the risks for this decision we are making, how will be the weather tomorrow, how many cups of coffee I am going to drink this month(I reckon that it will be many).
As you can see, many of the predictive analytics/models are linked to time, but they are not the only ones possible. In general, predicting refers to know the likelihood of a fact to happen before its occurrence. It is from here that machine learning starts to take place.
Not all the companies get to this stage of analytics due to its complexity and investments, even though it is a very important tool to make better decisions.
The main questions here is What do I need to do for the desired outcomes?
Prescriptive Analytics gathers all the previous questions (what is happening? why is this happening? what is likely to happen?) to build possible action plans for desired outcomes. From here, we can have the most perfect data-driven solutions ever (or almost that).
Artificial Intelligence is built upon prescriptive analytics by consuming a large amount of data to learn from it (going through all the previous steps) and make data-based decisions.
As popular and simple examples of prescriptive analytics outcomes, we have the recommendation systems that big companies rely on for their users: Netflix (movies and tv shows), Youtube(videos), Amazon (products), and so on. However, it can be way more complex than that when talking about business variables and constraints. Moreover, it is crucial to test the result models to ensure that they are providing meaningful recommendations.
We saw that each type of analytics is associated with one or more questions and the value of the data comes from the answers to them. Therefore, it is really important to make the right questions before diving into the data exploration. Otherwise, it can mean a waste of energy (and $money$) with no meaningful outcome coming from it.
And don't forget:
HIVERY trademark tagline