Machine Learning for Business Managers

Machine Learning is one of the hottest technology topics these days. With the proliferation of interconnections between humans and devices the amount of data that we are generating every minute is exponentially increasing and get value from that data is one of the main goals of many companies these days. It is clear that everybody sees the big technology companies that growth up analyzing and putting in value these huge amounts of data, as a reference and want to apply these techniques to their own business looking for ways to generate new line of business or increasing the revenue they are getting from the existing ones.

From a technical point of view, Machine Learning is not a new technology. Most of the algorithms that are driving our decisions today were developed 25 years ago. It is the increase of data, and the flexibility and power that the cloud brings to us the catalyst to this explosion. Apply Machine Learning algorithms is not a giant effort anymore so every single company in the world could do that this these days and get the benefits that it could bring to their business. That′s the reason why this is not just a technical discussion but also a business one, and we can see information, news and interviews in general or economy newspapers talking about Artificial Intelligence, Machine Learning Neural Networks or Deep Learning and the impact that it would bring to our lives.

Definitions

Artificial Intelligence is the machine (software) ability to reproduce the human behavior in any way. When Siri, Alexa or Cortana understand your voice, when Netflix recommends you a new Serie to view or when your smartphone camera recognizes faces are different types of Artificial Intelligence solutions.

Machine Learning is the use of statistical algorithms to automatically take decision in the way a human could do it, so it is a subset of Artificial Intelligence.

Supervised Learning, Unsupervised Learning, Reinforce Learning, Ensemble Learning and Deep Learning are different areas of Machine Learning to solve different problems using different algorithms.

The perspective of a business manager

From a business manager′s perspective, the value of this techniques would vary a lot depending of the type of business in first term, but also, in the data strategy that the company have in place these days. It is quite common to see companies where they cannot get real business value from Machine Learning mainly because they never did put attention to the amount and the quality of data they were generating from their business processes. There is no magic here, the value that you could get from a Machine Learning project is proportional to the love you give to the data. That is the reason why most of the work that your technical team needs to do is more related to Data Engineering (cleaning, transforming and managing data) than any other thing.

Step by Step

How I can start my journey into the Machine Learning arena? General steps to follow would be:

1.      Review your key business processes and try to understand the amount and quality of data you are getting. Put your focus on those processes where the quality of the data is better and get the momentum to improve your data strategy around the weak ones.

2.      Review the legal aspects of use this data, mainly if you are working with personal data, and ensure that you meet all the rules you need to meet in your business domain.  

3.      Run an Ideation Workshop where the business meets the technology and you can analyze which would be the first Machine Learning project you would like to run

4.      Identify the goals of the first project and the kpis you are going to analyze to understand the real value that this project would bring to the table.

5.      Let the technicians create the models but put them a red line: The model must be interpretable by your business team. The technical team must dominate the math behind the model, so they should look for a business way to show the results.

6.      Try it and refine it. It is not common to get the best results on your first attempt.

Skills for your team

You need Data Engineers, because you would need to deal with different data sources, and you would need to apply many transformations to the original data to get it ready for the algorithms.

You need Data Scientist, because you would need to understand the mathematical base of the algorithms when you need to choose the best one and try to

You need business users, because you would need to test the models against the reality and you need to put the results and way the model is going to take decision on business language.

You need developers, mainly because your model would be integrated into your business process and you need to interconnect it with existing solutions.