In recent blogs we talked about data mining and various types of analytics, such as predictive, advanced, exploratory, prescriptive analytics, and others. These are concepts that are best explained simply. In this blog, I will address the fundamental similarities and differences between them.
All of the analytics terms mentioned above have very specific and different meanings and usages, but they are all based on a common approach: Take a very detailed dataset, apply some mathematical algorithms, and find relationships within the data it contains.
The algorithms search for possible relationships and find the strongest ones to detect the most recurring patterns or rules in your past data. The most important set of rules is what we call a predictive model.
The algorithms can identify, as an example, that there is a strong relationship between a specific customer characteristic and the fact that the customer purchased a specific product, or that it churned. Or you can find that when customers purchase product A, they also purchase product B.
The predictive model not only tells you that a pattern exists, but it also measures the strength of this relationship.
You can answer various kind of questions following the rules detected. John MacGregor, in his SAP Press book Predictive Analysis with SAP, categorizes the possible business applications in finding trends, key influencers, segments, outliers, and associations.
- Trends help you understand how your business could evolve in the future.
- Key influencers explain what the underlying reason is of some events (why those customers are churning).
- Segments provide a categorization of your customers or products, or other items of interest in a format you can use to fine-tune your business (typically you would use it to create groups of similar customers for marketing campaigns).
- Outliers do exactly the opposite: Tthey tell you what items are very different from the norm and you can use them to see if there are issues (or if you are lucky, if whether there are incredible opportunities to take advantage of!).
- Associations highlight the previously relationships and let you use them in your business.
Once you have created the model for one of the uses above, you can do a couple of things: Apply it to new data or study the rules it contains.
When you apply the model to new data, you are really doing predictive or prescriptive analytics. As an example, in a list of prospects you find, based on past data, which ones are more likely to purchase a product. Thanks to the fact that the model measures the strengths of the relationships, you also know the mathematical probability of each prospect to purchase the product. You can now build an application for your sales people that will provide the daily list of prospects to contact, ordered by probability.
When you instead study the rules, you are doing more data mining, or exploratory analytics. Looking at the model structure, you can better understand the internal functioning of your business and decide to change it to better benefit the opportunities or decrease the impact of issues. For example, if the model highlights that you are more likely to be successful with downtown customers, you might want to centralize your stores and avoid opening new ones in the countryside or suburbs.
Just imagine what you could do by applying predictive analytics to your business. You may also be inspired by what other organizations have achieved. Discover the possibilities in our customer success stories.
Want more insight on data analytics? See 10 Myths About Predictive Analytics.