From Talk to Tangible. A Real-World Guide to Machine Learning.
Part 3: Exploration
In the last post, Education, I shared why Machine Learning is happening now, and the common organizational and technical elements that lead to successful Machine Learning initiatives. In this post, we are going tackle what I consider the most difficult and important phase — Exploration.
Challenges and Opportunities
In just about every survey I have seen, the top challenges to advanced analytics pursuits are the same — constructing a strategy, ensuring senior management involvement and designing organizational structure to support the initiative.
The underlying causes of these challenges tend to be simple to understand and many times straight forward to solve them. The first cause is the notion of developing a Machine Learning strategy, this tends to lead to a solution looking for a problem. The second cause is closely tied to the first where senior management perceives the Machine Learning initiative as interesting but are not committed to the initiative itself so their attention to the project dissipates over time. Finally, too often Machine Learning projects are performed in isolation from the ultimate users, which significantly increases the difficulty of change management as well as adoption.
Building Use-Cases
Successful Machine Learning initiatives have six critical elements — Use-Cases, Foundational Data, Workflow and Automation, Expertise and Tools, Agile Culture and User Adoption — when it comes to the Exploration phase most of the attention is on Use-Cases.
A Use-Case is how we directly connect Machine Learning as enabler to a corporate strategy or business imperative and how we ensure the initiative is sufficiently relevant and valuable to senior management and end-users. Identifying the right use-cases for Machine Learning is both challenging and rewarding. Challenging because we are dealing with a topic that is conceptual and not well understood, while it is also rewarding because there is an opportunity to solve intractable problems.
Identification of business imperatives is the first step to building a strong use-case. A business imperative is a catchall that is a tangible business opportunity or business problem or both. Examples of business imperatives can be found below. A business imperative represents something that has broad agreement, has a sense of urgency about it and fits into a Machine Learning domain.
After reviewing the examples, it may have occurred to you that these are common business things that many companies contend with and that’s the point. When collaborating with senior management during the Exploration phase we spend very little time on the topic of Machine Learning and most of our time detailing exactly how to solve for the business imperative.
A use-case requires a few questions to be answered, for example, Life-Time Value
Why do you want to predict LTV?
How is LTV defined?
If we were able to predict a customer’s LTV with underlying factors how would you act on the information?
Who are the primary users and how and when would the need to receive this information?
How would you measure if the LTV Machine Learning initiative is meeting expectations?
Ideally, you develop a portfolio of use-cases and have senior management prioritize the use-cases to ensure alignment. Also, important to note, from my experience, you can conduct a Proof of Concept with a single executive sponsor but to operationalize the Machine Learning initiative you will need at least two executive sponsors.
Where We Can Go From Here
Once you have the use-case identified, you will need to validate it through a Proof of Concept or more accurately Proof of Value because value is what will ultimately determine if the Machine Learning initiative proceeds. In the next post, I will describe the key elements to a successful Proof of Value.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — -
David is the practice lead for Advanced Analytics and Data Visualization for Slalom Philadelphia. Slalom is a modern consulting firm focused on strategy, technology, and business transformation.We help companies tackle their most ambitious projects and build new capabilities. @slalomphilly
