Evaluating ML Opportunities, Part 1

Ask These Questions First

James Kotecki
Machine Learning in Practice

--

How should organizations best prioritize their ML projects? What makes for a good project versus a bad one? Since machine learning is new to many organizations, these are relevant questions even for experienced executives.

This post will lay out key Prerequisite questions to help you decide how to move forward. In future posts, I’ll cover Project and Process questions. But of course, Prerequisites come first. While you don’t have to have concrete foresight about the entire project, you should at least be able to answer the questions below with “yes.”

Be honest with yourself and with your team — the inability to answer these four questions is a good sign that, no matter how interested you are in machine learning as a concept, you’re not ready to do machine learning in practice.

If that’s the case, don’t be discouraged. It may take hard work, but once you’re able to answer all of these questions, you’ll have built a solid foundation for machine learning that will help you succeed. It’s not as if you can rush through these questions to “get to the real machine learning.” Without these components, you’ll eventually get stuck.

Do You Have Project Ideas?

Some leaders are interested in machine learning but have no idea where to start. You don’t need fully-baked ideas to get started, but you should at least have the equivalent of dough.

A good place to begin is often with your data. Ask yourself how you are using your data today. Could you automate any of those processes? Would you like to find new insights in the data that your people lack the time or ability to uncover?

Machine learning has so many different applications across industries and data types. Keep in mind that to be effective, your ML ideas need not be original or feel especially creative. It’s ok if dozens or even hundreds of other companies have used data to automate a process and save money. That means you can do it too.

You don’t need total clarity to move forward — the goal here is just to get you to the starting line of the evaluation process. At the very least, you need a sense of what you want to do. You also need a sense of your data, which leads to the next question:

Do You Have Access to Data?

Even if you have machine learning project ideas, they won’t go anywhere without data to support them. It’s ok if you lack perfect clarity about the precise data you need. At this point, even a belief that you can get it from somewhere is enough to start a more in-depth evaluation. If you don’t have the data internally, you may be able to procure it from external sources.

Data and project ideas go hand in hand when selecting machine learning projects. If you have neither data nor ideas, your organization is probably not ready for machine learning at this point.

Do You Have Data Science Resources?

Do you have access to people, either internally or externally, who are technically capable of executing a machine learning project? You’ll likely need not only data scientists but engineering talent to pull this off. You may also want to confirm with your technical team that they can access the tools/hardware they need to do the job (knowing that the shape of that job might be fuzzy at this point).

Do You Really Need Machine Learning?

There are many problems that might seem to require ML, but in reality, can be solved with statistical techniques as straightforward as linear regression. Don’t over-engineer a machine learning solution for a simpler problem.

Simpler ideas may still be very much worth doing, of course. You may even want to do these projects first, before jumping into ML, precisely because they are easier and can serve as a proving ground for underlying data readiness.

If you can answer yes to all of these questions, you’re ready to begin evaluation of your specific project. I’ll cover those questions in my next post.

James Kotecki is the Director of Marketing & Communications at Infinia ML, a team of data scientists, engineers, and business experts putting machine learning to work.

--

--

James Kotecki
Machine Learning in Practice

VP, External Affairs for Agerpoint, a spatial intelligence platform for crops and trees. Also a talk show host for CES.