The 5 Types of Data Analytics: From Hindsight to Foresight
BI Connector Team |
Businesses today breathe Data Analytics for decision-making.
Almost all organizations (be it small, mid, or large-sized), have already built a solid dashboarding setup to visualize their existing dataset in a hindsight fashion.
A lot of these organizations perceive Data Analytics as a one-time process of building dashboards using the following 4 steps (with some enhancements and fine-tuning from time-to-time):
- Building connectivity to the different data sources
- Performing the necessary ETL/ELT jobs
- Creating a single source of truth
- Developing/sharing dashboards for decision-making
After the completion of this perceived one-time process, the leadership teams usually begin to feel that their organization has hit the saturation point of the Data Analytics practice.
However, that’s just the starting point and provides the basic setup required for your data analytics initiatives. There’s so much more!
Types of Data Analytics
When we understand the different types of Data Analytics, we can easily recognize that it is much more than that one-time process!
In this blog post, we’ll take a quick look at the following types of Data Analytics:
- Descriptive Analytics
- Diagnostic Analytics
- Predictive Analytics
- Prescriptive Analytics
- Cognitive Analytics
Quick heads-up: The above list is primarily based on the Gartner Analytic Ascendancy Model, which was released back in 2012. The last type, Cognitive Analytics is a recent addition to the list. There’s no surprise if this list expands in the future, given the innovations happening in the Business Intelligence (BI) and Data Analytics landscape!
Let’s see each type in detail below.
In this type of analytics, you visualize the data to understand what happened so far. In other words, you’re giving life to your existing data with some visualizations to understand what happened in a hindsight fashion.
The decisions made based on this type of analytics are purely reactive in nature.
A quick example is the comparison of your actual vs target sales for a particular time period, say last quarter. The target is shown as a benchmark in your visual, and your actual is shown to compare both.
If the sales exceeded the target, you can conclude that your sales team is performing well. If it’s below the target, you have 2 choices –
- Make a random reactive decision to shuffle the sales team and see if the new team achieves the target next quarter on a trial-and-error basis
- Choose to figure out why the sales targets aren’t met and make a data-driven reactive decision, which takes us to the next stage of Analytics, which is Diagnostic Analytics
In the first choice, the decision-maker relies on luck (which hardly works in a heavily competitive business world), while in the second one the decision is to perform a Root Cause Analysis (RCA), which takes us to Diagnostic Analytics, the next stage of Analytics.
While Descriptive Analytics gives a picture of what happened, Diagnostic Analytics seeks to answer the question – why it happened.
In the same sales actual vs target example, let’s take a step further to analyze why the sales didn’t meet the target.
In this step, we’ll need to consider several other factors that are closely coupled with the sales. In this example, let’s consider the following factors:
- Each salesperson’s performance
- The market demand for the product
- The quality of your Marketing Qualified Leads (MQL)
You can add many more dimensions to the above list based on the factors like climatic seasons, festivals, regional laws, etc that could affect your product sales.
When you correlate the quarterly sales with the above factors in that quarter, you can quite easily conclude which factor(s) caused the sales to drop.
Based on the insight, you can make a decision to make improvements in that particular factor to achieve the sales target in the upcoming quarters.
It is easy to perform Diagnostic Analytics with modern data visualization tools such as Tableau and Power BI, with their interactive features and drill-down capabilities.
If you’re using OBIEE and struggling to unlock data insights, you can try connecting Power BI or Tableau to OBIEE or OAC or OAS (for self-service data visualization) through BI Connector.
Predictive Analytics, as the name suggests, is about predicting a result in the future, based on an analysis of your past data.
From this stage, it is all about foresight, in order to make decisions for planning appropriately ahead of time.
The probability of getting an accurate prediction highly depends on the size of the dataset. The bigger the dataset the higher the accuracy of the prediction.
In certain cases, it is recommended to remove the outliers in the data if they were a result of an uncommon scenario.
Now let’s apply Predictive Analytics to the same sales data, as a simple example.
I want to know what my actual sales will be one year from now, based on my past sales data.
In this case, I can use a Simple Moving Average(SMA) of the past x years’ sales data to forecast the sales in the next year.
SMA is a basic and simple example. When you bring multiple dimensions like the product, geography, etc into the picture, you can predict sales using advanced calculation methods and Artificial Intelligence (AI).
Prescriptive Analytics is all about providing decision suggestions to achieve a desired outcome in the future.
Prescriptive analytics is a great way to evaluate the possible options for achieving the desired outcome.
An example of the desired outcome could be an increase in revenue by x% or a reduction in operating cost by y%.
The suggestions are made through advanced analysis powered by AI and Machine Learning algorithms.
The decision-makers can simply select the best recommendation that’s already driven by data rather than scrutinizing the data and identifying the available options manually.
However, prescriptive analytics depends highly on how well you train the AI to understand the data in order to uncover patterns and trends that humans couldn’t identify easily.
Cognitive Analytics, as the name implies, is to mimic human thinking into data analysis.
For example, when a human views the data of year and sales, he/she can recognize the year column, though is a number is not a measure but for identification in terms of time, while the sales column is a measure.
Humans can understand the relationship between different columns in a dataset. When computers mimic the same kind of thinking, cognitive analytics comes into life.
AI, ML, and Deep Learning (DL) also power cognitive analytics. The Natural Language Processing (NLP) capabilities in leading BI Platforms (such as Power BI, Tableau) where the decision-makers ask a simple question, like “What’s the sales revenue in the last 5 years” and get the answer as the response is a simple example of Cognitive Analytics.
Cognitive analytics can also be used to drive automated decisions, especially in unforeseen scenarios.
Moving from hindsight to foresight and automating decisions isn’t an easy task. It involves a lot of trial and error, and steep learning curves for the data scientists and developers. However, the perceived benefits are the motivating factor for the organizations working on the advanced stages of analytics.
Subscribe to BI Connector
Get the latest BI Connector news, articles, and
resources, sent straight to your inbox every month.