How Machine Learning Could Unlock The Potential In Data You Already Have

Is machine learning only for companies with endless resources? Companies like FANGs (Facebook, Amazon, Netflix & Google)? It may surprise you how accessible the benefits of machine learning are for the company.

Businesses have already collected a valuable harvest of data. Few people realise that with machine learning, this can drive businesses to even greater heights.

What exactly is machine learning? How does it differ from artificial intelligence?

Machine learning (or ML) is an approach to artificial intelligence.

Classic artificial intelligence design is a step by step instruction algorithm. It guides a program through every eventuality that the program may come across. With machine learning, the program doesn’t need to be explicitly programmed. It is designed to learn and re-calibrate itself.

Let’s take an example: Email spam filter. It may be programmed to move emails that contain over a certain number of images into the junk folder. This is where machine learning comes in. A machine learning program will learn how the user interacts with some of the emails marked as spam. It will then train itself to categorise future possible spam with more accuracy. A classic spam filter design would follow its instructions without learning new insights.

How long has machine learning been developing?

Machine learning has been around since the 1970’s. But the horsepower required to run ML programs was prohibitive until recently. We now have the architecture to make machine learning accessible to small and medium businesses while refining and reworking sophisticated variations on algorithms that have been around for a long time.

What can machine learning do for businesses? Is it worth it?

The benefits of ML are as varied as its implementations. Caterpillar Marine is a company that builds and runs hardware and software for ships. They used Internet of Things sensors to collect vast amounts of data across their fleets. They then used machine learning to analyse the data and gain insights. The resulting improvements saved them millions of dollars per year. This was money that was being wasted until they adopted ML.

On a more intimate end of the scale, there’s Ecree. Ecree is a writing assessment program that evaluates essays for students. In under a minute it gives clear feedback on how the student can improve. Ecree uses 36 metrics to help analyse and develop writing skills almost in real time.

How can you start to reap the benefits of machine learning?

Define your goal

First you need to define the goal of the ML implementation. Then you can assess whether machine learning is the appropriate tool for the job.

It helps to look at some of the most common implementations of machine learning. In general this is either:

1. Automating workflows

2. Detecting patterns in large data sets

3. Helping with customer interaction

For example, an ML program can learn to search vast numbers of incoming tweets. We train it to detect negative customer experiences and flag them for a closer look. A customer service operator can then deal with the issue. When a company reaches a certain size, this saves a lot of expense on large teams.

Or a program can learn to detect fraudulent credit card transactions. We can train it to develop an increasingly sophisticated understanding of what to flag for closer inspection. We can train similar algorithms for insurance companies to streamline the claims process. Both of these examples save both expenditure and time.

Then there’s speech recognition. Programs like Siri and Alexa, which use machine learning for voice recognition. Programs like these are feasibly accessible for the first time. Retail commerce uses ML to tailor customer experiences. We can train recommendation engines like the ones used by Amazon, Spotify and Netflix.

Assess your company

Evaluate whether you are ready. Machine learning relies on having a large quantity and quality of data. “Garbage In, Garbage Out” applies especially to ML.

Widespread adoption of Software as a Service (SaaS) means that most businesses have access to enormous amounts of rich data . We can use these to train machine learning programs. You may be sitting on a gold mine of data that can be used to transform processes with machine learning. You’ll need to work with experts and increase staff knowledge in your company. This can make or break a successful machine learning adoption.

Assess your Industry

Is your industry ready? If you adopt machine learning, will it have a significant impact?

Again, a good indicator here is playing to ML’s strengths:

Automating workflows, enhancing data driven decisions and helping with customer interactions. These are the areas where machine learning most commonly has a good return on investment.

Assess the competition

Are they implementing machine learning? If so, that may be a good sign. Find out what edge they are gaining from it, then get creative on how you can go further. As well as direct competition, look at parallel industries.

Predict and evaluate

Cost is a key factor. Can you predict a worthwhile ROI? In all probability there are areas that machine learning can help you. If you prepare for a healthy adoption process, you could experience significant benefits.