Understand the 4 Types of Data Analytics
As you put your toe into data analytics, it may be puzzling at first. This mishmash of technical concepts and algorithmic theories is a pain to digest. Still, it’s better to understand what you are going to do before you are actually doing it.
In the previous piece, we talked about some of the fundamentals of data analytics. So now let’s dip your toe a little deeper.
In our view, data analytics is an umbrella term for 4 related analytical techniques.
- Descriptive analytics describes what happened both in and out of your business.
- Diagnostics analytics takes a step further to explore why those things happened the way they did.
- Predictive analytics, built upon the above two, attempts to forecast what is going to happen.
- Prescriptive analytics, taking into account all the information at hand, suggests what to do.
It is clear that descriptive analytics and diagnostic analytics, when combined, help to narrate the past, whereas predictive and prescriptive analytics navigate the future. In fact, a stacked Venn Diagram may best illustrate their relationship.
In this setting, descriptive analytics is the basis for diagnostic analytics. These 2, in turn, underpin predictive analytics. And once your analytical framework is solid enough, prescriptive analytics can be added as a final building block.
Note what the 2 axes got to say: the further you go with your analytics, the messier it gets, yet the more benefits you can earn.
As we talk to data scientists or read their papers on analytics, often we would be bombarded with jargon and theoretical concepts that may require extensive training to grasp.
Here, descriptive analytics is a different case because it is merely a way of reporting - nothing more and nothing less. And it’s no data science. You know your business is doing descriptive analytics if you periodically collect and visualize data.
The end products of this process could be as dull as a report (such as your sales record or P&L statement) that is manually filed via a spreadsheet. Or it could be a little more intuitive as a graph or dashboard. Perhaps you can tell by now, most Business Intelligence tools that permeate the business world these days fall into this category.
Descriptive analytics helps to answer the “what”, “how many”, “when”, and “where” of variables. And addressing these plain questions provides the required data for further analysis. This makes it the baseline of your whole analytical strategy.
Let’s say you own a flower store. Running this kind of business is tough because, on one hand, you are subject to wild seasonal fluctuation and, on the other, the life-cycle of your products is extremely short - your flowers could not bear for long. So, first thing first, you want to stay on top of practically everything: from sales, inventory, to your suppliers, and your customers.
Having heard that descriptive analytics might help, you decided on a basic BI tool. Though you could not ask for more from what you have paid, the tool still does a fine job collecting, retrieving, managing, and visualizing your critical data.
Getting to know what happened in your business is all-important, but it’s not enough to help you draw valid conclusions and make informed decisions. This is why diagnostic analytics is spearheading the strategy development in large companies, where top executives oftentimes demand the justification for every figure on the reports. This approach seeks to examine historical data to give a clearer picture of the past.
Find out about our Data & Advanced Analytics services, and how we can help accelerate your data journey.
Doing diagnostic analytics call for the use of some advanced methods such as drill-down, queries, data discovery, and data mining. Sophisticated as they may sound, their goal is fairly simple. By assessing the correlation between 2 or more variables, they help you answer some questions that could not be answered by staring emptily at your reports.
For the sake of clarity, it may help to get back to your fictitious flower store. Recently you saw an uptick in sales and an increase in your website traffic, which is odd because you were not doing any kind of promotion at the time. In quest of an answer, you decided to use some diagnostic methods. By examining both internal and external data sources, you found out that there is a sudden increase in the number of weddings held in town during the past few weeks, which is highly correlated with the hike in your sales and web traffic. Still curious, you did a couple of searches and learned that last month is one of the most popular ones to get married. Your puzzle is solved.
Predictive analytics attempts to forecast what might happen. It is no magic actually. In its essence, predictive analytics takes the findings from descriptive and diagnostic analytics to build a model of the past. A model could be thought of as a reproduction of what had happened. You predict by extending this model to the future, assuming that things would be the same as they were in the past. This way of forecasting is known as extrapolation in the parlance of statistics.
To make your predictive analytics tick, you may need to employ regression analysis, forecasting, multivariate statistics, pattern matching, predictive modeling, and forecasting.
It’s time to revisit your floral business. The end of the quarter comes near and you need to update your sales forecast. This is critical because, without a definite plan in place, your business would be as short-lived as the flowers it sells. So by looking at the past sales record, you can tentatively set up a predictive model saying that sales for next month would be this much, for example. You could then stock your shelves accordingly so that supply would meet demand, and waste is reduced to an acceptable level.
Because your business is constantly evolving and the market constantly changing, you need to iteratively update your model.
People often draw an analogy between predictive analytics and the crystal ball, which is completely not true. Or at least we are still a very long way to the point where analytics could allow for such a low margin of error.
Prescriptive analytics guide future actions - it prescribes what to do. Given all the possible models and scenarios generated by other analytical techniques, it does a multitude of algorithms to suggest a course of action that may yield the most desired outcome.
The advent of big data, artificial intelligence, and machine learning is what enables prescriptive analytics. The reason is that they help capture and process a vast amount of data that is beyond humans’ capability.
So your floral business is doing far more than expected, people start to order your flowers via website and phone call. It may be too much for a small store, but you want to do route optimization to reduce costs. This is where prescriptive analytics does its job. So every time there is someone ordering, it would build as many scenarios as possible. Then by doing some math tricks based on predefined rules, it would suggest the one route that saves you the most time, money, and energy.
If you want to learn more about data analytics, read the report “Driving toward analytical ubiquity with embedded analytics” to discover how data analytics evolved over time, and which strategies top companies are using to approach data analytics.