How Can Companies Benefit From Machine Learning?

Machine Learning (ML) is one of those technical subjects that keeps on being mentioned as a business solution. It is often confused with Artificial Intelligence (AI), so it’s worth stating here at the start that there is a difference.

AI is the simulation of human thought or behavior and ML is the ability to learn and draw conclusions from data without being explicitly programmed to do so – the machine can just look at the data and start drawing conclusions.

The two subjects are often connected because ML can be used as a subset of AI. You build a system that analyzes data and learns from it and this can then trigger the AI to make certain decisions. Connected, but not quite the same.

A good example of machine learning is the ability to learn from images

Type ‘dog’ into Google image search and you will see millions of photos of dogs, but Google doesn’t need to only use images that were labeled as a dog by whoever uploaded it. By training the system they can feed it a large number of images that are verified as dogs and then the system can learn the distinguishing features, allowing it to recognize an image as a dog in the future even if it has no further information – in the same way, a human can recognize a photo as a dog even if it is a photo we have never seen.

This all sounds a bit complex and mathematical and as you might expect, it is complex. If you follow the ML community online then you can see arguments raging over complex subjects that the rest of the world cannot understand. However, a general understanding of the possibilities of ML is important for business leaders because this is going to create several opportunities in many industries. It is getting easier to deploy these solutions as well. Companies like IBM with their Watson product offer ML solutions in the cloud – you don’t need to build anything, just pay as you go.

Every executive knows that their company has a huge amount of data on customers. You know what they purchase, what they like and dislike, and even how they behave on different days of the week. By applying ML to this information it should be possible to accurately predict the behavior of customers and to personalize the experience you give to them.

Here are a few more specific ideas of business solutions that can be powered by Machine Learning:

1. Customer lifetime value modeling:

who are your best customers? What differentiates them from others? Are there ways you can nudge good customers so they become great customers? Learning about the behavior of customers over a long period of time can yield incredible insights into how you can serve them better.

2. Churn modeling:

many subscription businesses rely on customers paying a bill every month – video streaming, Internet, mobile phone contracts. If you can analyze customer behavior and predict which ones plan to leave your service before it happens then that’s extremely powerful and can allow you to offer a deal that may retain the customer.

3. Dynamic pricing:

industries such as airlines and hotels can learn from past behavior to dynamically price their services, rather than only ever using fixed rates. Don’t leave cash on the table when your customers are prepared to pay more.

4. Customer segmentation

can you categorize your customers using finer granularity, rather than the very broad use of demographics or location? What does it reveal? Can you learn from one group of customers and apply those lessons to others?

5. Recommendations and special offers:

linked to churn modeling, but not restricted to subscription businesses. If you know your customer always shops at the end of the month and likes a particular category of products then you can help to trigger purchasing activity with highly personalized special offers or recommendations that are targeted at the individual customer.

In the customer service environment there are also many applications for ML to work on predictive support. It’s possible to listen to a conversation between a customer and support agent in the contact center and for the system to be presenting the agent with the most likely solution to the customer’s problem – before they even need to search for the information. It just pops up automatically.

I have also seen banks and other major companies use ML to tag and categorize customer questions with solutions. This allows the system to build an automatic knowledgebase over time that contains every question a customer has ever asked, with the correct solution. This is extremely useful and can then be connected to a chatbot or used by a human to help the customer.

It’s clear that although ML is a complex subject, it does have some very real and useful business applications. Cloud-based services are making it easier to experiment and deploy these solutions and I’m sure that it will feature as an important technology trend in the 2020s.