Data "Witchcraft", from crystal ball to watson analytics
- But the underlying idea was and remains valid, we may be able to respond to:
- What are the parameters that most impact a customer's decision to buy one product or another?
- Could we predict the malfunction of a particular system?
- How do they impact behaviors outside my organization, my business?
In classic business intelligence systems, the answers to the above questions are obtained on the basis of predefined systems, in which we have already identified our market target, we have a maintenance plan of the system manufacturer or there is a market study carried out by a third party...; that is to say, we have predefined the answers and we simply check their level of compliance.
But what if the answers weren't the right ones? Then we would be measuring indicators that do not reflect the reality of our business.
Fortunately, mathematics, and in particular statistics and statistical models, describe techniques for dealing with these complex problems that may not have been available to everyone. Until today.
Watson Analytics, comes to cover that space of self-consumption in predictive analytics. It is a service offered by IBM (with a very complete free version), which offers relevant information thanks to the analysis of massive data capture. To be specific, Watson is capable of detecting associations and relationships between data, offering the user visual answers about future trends. In addition, the program uses a natural language, so it is possible to launch specific questions that will be answered in a clear way: what are the sales that will close this month, what unforeseen expenses will we bear the next quarter? which customers will cause downtime the next 6 months?....
But also with Watson Analytics we can:
- Obtain quality information thanks to published data, will we be able to obtain intelligence from them?
- Search for Predictive Models and get results like:
- Representation of predictions by means of a dashboard
I invite you to try it, there are test data sets that can help you and... predict!