Is machine learning always the right answer?

3 things you need to consider before diving into ML…

Machine Learning is one of the most in-demand skill sets in today's market, but with demand for Data Scientist outweighing the supply available, is applying machine learning always the right answer?  

As a data science head-hunter I am always hearing the buzzwords ‘machine learning’ flying around, and that as a data scientist machine learning should be your first language - if it's not, then can you really call yourself a data scientist?  

This blog aims to shed some light on whether machine learning is always the right solution and if not, what should be? 

 First, let’s remember what data science is used for and why it is needed.  

• Data science allows companies to make better decisions; improve and develop a product or service; identify and act upon opportunities such as opening new revenue streams and identifying trends to stay competitive.  
• Machine learning, however, enables data scientists to build models which can automatically learn from itself. Machine learning has always played a big part in data science but only recently has been the hot new topic.  

So, should it be in every aspect of data science, and can a data scientist be a data scientist without it?  

Diving into machine learning without really knowing what it is and how it can improve a business can be amiss. If we hire a machine learning guru will this guarantee success? The mathematical nature of machine learning can be very daunting. So, the question is whether machine learning can benefit your business or you as a data scientist. This all depends on the problems you are trying to solve and the data you can collect.  

So, what should you consider when thinking about employing machine learnings, here are 3 key questions you should ask yourself. 

What’s the business problem?

If your purpose is to build recommendation engines, then machine learning is the right answer. If your purpose is to solve complex business problems that can be used through statistical modeling, then machine learning is not your answer. It is important to ask yourself what are you trying to achieve and how will applying data science get you there. Figure out what you want to do and why and what method is the best answer (ML isn’t always the answer, but it could be). So, what does this mean for you? Well if you are a data scientist and have been tasked on how to increase sales through data, it's Important to think is machine learning going to get you there? Is there a faster, more efficient alternative which can get you these results? Only after answering this question can you decide whether ML is the right answer.  

What methodologies are needed? 

Data Scientists are typically hired to solve a problem that can’t yet be solved. CEO’s and senior stakeholders will want to hire data scientists to solve these problems by whatever means necessary. More and more business functions and companies are looking to machine learning to find the solution. But should they be considering other methods such as linear regression, logistic regression, bayesian methods, random forest trees, time series, decisions trees. These methods are much simpler, quicker to implement and can be easier to solve the problem. However, the common mistake is 'let's just implement machine learning because Google are doing so'.  A lot of time can be wasted if the answer to the questions is “can machine learning solve this problem?” The answer could well be a 'no' or it could be a 'yes', but it’s important to look at what is needed and what your methodology should be.  

Why business analytics focused data science is crucial? 

More and more companies are viewing data science as a tool to solve complex business and product problems. E.g. how do we increase user growth? How do we increase ROI and how can we improve decision making? This is when conceptual thinking and thinking how to solve the problem is more important than the techniques that we can practice. That doesn’t mean that machine learning techniques cannot be used to help understand the important patterns within data. Many people who stress machine learning tend to rely too heavily on it and statistical tools, rather than trying to understand the narrative and casual interference underlying the mechanisms that created the data. 

 The ongoing conversation between using machine learning to solve complex business problems appose to statistical modeling or deep dives is ongoing and an interesting argument. I’d love to hear your thoughts and what you think? 

If you have any questions or are a data scientist looking for a next role or a data-driven company looking for your next data scientist, then please get in get in touch on [email protected] or you can reach me on LinkedIn (Alex Howman)

If you want to receive more news and insights from the CN team remember to follow-us on LinkedIn and Twitter!

Alex Howman, Data Science Head-Hunter at Campbell North