Blue Dots Partners

This is the second of a three-part series on machine learning. See Part One here. In the first article, we laid out the basics of machine learning. In this article, we will make the case that business leaders should start machine learning initiatives immediately. In Part 3 (next month) we’ll construct a framework for integrating intelligent machine thinking in your business and explore how to align these disciplines with your growth objectives.

For centuries, human beings have used devices to aid in the exploration, observation and cataloging of the world around us. We used telescopes to gaze into the heavens to observe the paths of stars, moons and planets. We used microscopes to observe the tiny world of single-cell entities.

While we continue to explore the physical universe, another world is exploding – the universe of digital knowledge, communications, images and numeric data. We can’t see it, but you likely have a sense that it is proliferating rapidly. It is estimated that the human race generates 2.5 quintillion bytes of data every day and that 90% of the world’s data has been created in just the last 2 years.

To explore this world, we use computers and machine learning. Machine learning has already been applied to recognize the content of unidentified images, detect cancerous cells, extract the meaning from terabytes of collected tweets and Facebook posts and more. As we explored in the previous article, machine learning is no longer new or novel. It is time to put it to work in every business.

Here are six reasons why your business should devote energy and resources to a machine learning project today:

Reason #1: New insights that can drive growth await!

Most businesses spend a lot of time obsessing over products, customers and sales. If you get these right and you’ve taken care to assemble a great team that is aligned with a clear strategy and a positive culture, you are well on your way to success.

Unfortunately, this perfect picture is seldom the reality or the reality is short-lived. Products change. Competitors change. People, strategies and financial conditions change. Machine learning can help detect and anticipate business changes just by framing a few simple questions. Today, there are five key questions machine learning can answer:

1. Is this A or B? (or C or D or E?)

What if you could learn which of your new customers was likely to be a high-volume buyer within the first 30 days of working with them? Machine learning can look at data and patterns from your existing customer base and help you classify your new customers early on. By mapping your marketing and “good customer” activities to these new customers at the outset, you can increase the value of the business relationship.

2. Is this weird?

A classic context for the “is this weird?” question is credit card fraud detection. Using past purchase histories, the location, the store, the time of day and descriptions of items being purchased, credit card company computers attempt to detect in real time when a particular transaction is likely to be fraudulent.

Could your business benefit from a similar approach? Perhaps you’d like to be alerted when a particular customer changes their buying habits. Or if you are in the delivery business, would it be helpful to know when a vehicle is not where it should be? These are insights that a machine learning initiative can produce.

3. How much? or How many?

Want to get a sense of how many units your customer will buy in January? What is the best predictor of January sales? Is it the weather? Oil prices? Past purchases? Analyzing combinations of related data sets to reveal relationships is perhaps the oldest form of insight. Machine learning can tirelessly parse through myriads of data sets to help you forecast and predict future business events.

4. How is this organized?

The previous questions all presume you have some knowledge of the relationships embodied in your data. You know, for example, that your customer’s purchases are likely related to product features, price and satisfaction. But what if you had data where the relationships amongst the various pieces of data were unknown? Consider a random collection of data like your website’s log of clicks, page requests and downloads. What does it all mean?

Special algorithms exist that can help group and cluster raw information into meaningful clusters and illuminate the patterns. Using these new-found insights, perhaps you’ll discover a pattern where visitors get stuck on your site!

5. What should I do now?

The final opportunity for new insights is to look at your business as a series of decisions. Algorithms that “self-adjust” based on the situations they encounter are the essence of self-driving cars, warehouse robotics and recommendation engines. Every time a consumer downloads or streams a movie, Netflix’s machines are using that information to update its recommendation engine.

Perhaps your business has a similar opportunity. Could machine learning better guide your customers to the right solution (based on how other customers have consumed your products or services)? Could machine learning help you choose when to replace or repair a delivery vehicle?

If any of these five types of questions are relevant for your business, you have a reason to start a machine learning initiative now!

Reason #2: Implementing machine learning is good for your business health

Another important reason to embark on a machine learning initiative is that the process itself will catalyze healthy changes in your business.

Asking questions

By simply asking your team one question – “What do we wish we knew that, if we did know, could advance our business?” – you may expose information gaps that are limiting your business growth. At the very least, you may discover that one department is asking questions that another department already has answers to! In any event, the discovery of which questions need answers can be a key learning in itself.

Data-driven decision making

Basing decisions on evidence, not opinion, is a mantra that has been around for a while and is also core to the Blue Dots philosophy and methodology. But when it comes to complex organizations and multi-faceted decisions, evidence is sometimes hard to come by. Because computers and machine learning can process multiple facets of your business simultaneously, you can bring data to bear on difficult, complex decisions.

Revealing the gaps and silos

When you embark on a machine learning initiative you must have two ingredients: a question worth answering and specific data that might help answer that question. Often times, with a question in hand, the first realization is that the data you need isn’t being collected, or it is in bits and pieces around the organization, or it is not organized in a way that is particularly useful. Implementing a machine learning initiative might just be the forcing function you need to align your business processes with your strategies and to begin collecting the right information in a useful form.

Any or all of these benefits are reasons why implementing a machine learning initiative can be good for your business health.

Reason #3: Competitive advantage will be won or lost

Your competitors may already be developing the capability to quickly parse information and gain rapid insights about their customers. If your competitors are better than you at effectively turning these new insights into superior products and targeted sales strategies, they can become a major threat to your business.

Alternatively, if you start now to develop these capabilities, your organization will have the building blocks to create competitive advantages. Starting immediately might well be the catalyst to a winning strategy!

Reason #4 – Catching up gets harder

In Olympic challenges, the starting gun goes off and the race begins. In business, maybe there’s a starting signal and maybe there’s not. Maybe your business found no reason to start the race but others already have. The reality is that the race for adoption of machine learning to solve humanly-impossible challenges has indeed already begun.

If you haven’t embarked on a machine learning initiative yet, your business is already behind the curve. As it is now in the common vernacular, machine learning is attracting start-ups and investments. More research projects are getting funded and more people are experimenting with machine learning. When this happens, new tools and techniques soon follow. These tools and techniques then grow in sophistication and capability. If you don’t start applying machine learning to your business soon, catching up will get harder and harder.

Reason #5: Time is your enemy

While the steps to gain insights into your business are becoming increasingly straightforward, it may take significant time for your organization to produce and harvest these insights. You will need time to:

• decide which question(s) are the most important ones to answer
• identify what data you have and what data you need
• hire a data scientist
• assemble the talent that will support your machine learning initiative – from sales, IT, analysts, product management, manufacturing, etc.
• align your newly assembled team (and re-align the teams from which they originated) with the goals and ambitions of the initiative

Further organizational changes may be required to implement the insights. Perhaps your machine learning initiative will reveal some low value activities that should be discontinued. You may want to eliminate or reduce emphasis on these activities and ramp up higher value activities to implement the insights gained through machine learning.   Training or hiring may be needed in order to obtain the skillsets needed to implement the new insights. This too takes time.

The bottom line is that weeks and months may be necessary to begin harvesting the benefits of machine learning. Learning takes time and the sooner you start, the sooner the learning begins.

Reason #6: Machine learning tools abound!

As little as two years ago, a business might legitimately have claimed that a machine learning initiative would be difficult and a bit clunky. The tools to produce insights were relative axes when a scalpel would have been a preferable tool.

Those arguments are no longer valid. Algorithms, tools and techniques for machine learning are pouring out of the labs and into the hands of many. These resources are abundant and often free to start.

IBM’s Watson Analytics and Microsoft’s Azure Machine Learning Studio are powerful and extremely easy-to-use tools that allow you to upload data, marry that data with other data sets and, using standard algorithms, quickly get insight into your data. In 2015, Google open-sourced its TensorFlow™ software library for numerical computations. It was the result of five years of private development and is now freely available to the public. Salesforce’s Einstein project recently took first place at a machine learning and comprehension competition at Stanford where it competed against both IBM and Microsoft. It promises some ground-breaking technologies with respect to processing large amounts of text, structured data and images.

The point is that the barriers to starting a machine learning initiative are the lowest they’ve ever been. Tools, storage and resources are abundant and there is no reason to wait for something better.

Conclusion

There are at least six reasons why your business needs to start a machine learning initiative now. It really doesn’t matter what aspect of your business you choose to focus on for a machine learning initiative, so long as you harvest what you learn from the implementation and apply those learnings to the next machine learning project. In the third article in this series, we will explore how to select and implement a machine learning initiative.