Machine Learning for Product Managers: Defining the business problem

Problem of Data — it is abundant!

Every company is overflowing with data. They look around and see innovation is happening in the industry. Executives hear from their customers about their AI strategy.

Management sees competitors with AI solution and make critical moves that bite into their addressable market. With all this background noise, the immediate reaction for the management is to conclude that we got to do something with our data and let us go and hire some data scientists. That is no different than ten years ago when mobile was very hot, and the thought was to hire a mobile developer who will magically launch a mobile app for the company. You are assuming that Machine Learning is the solution and are looking for a problem it can solve.

What is the problem we are trying to solve?

Machine Learning is not magic. Machine Learning is a solution. One has to define what is the problem we are solving clearly.

Beware of the temptation to use the hot technology like Machine Learning to look for a problem it can solve. Or finding a problem without asking the question if that problem is big enough to tackle and invest resources. Machine Learning is like a drill bit. Which drill bit you use would be dependent on the problem you are trying to solve.

It’s important to define the problem you are trying to solve, the business results you are looking forward to achieving and the benefit you are trying to find for your customer. Once the problem is clearly defined, then one can start thinking about the data we need, the model to create, the algorithm to use, the insights and action to take based on the predicted insight.

Product Manager Collaboration

Defining what problem to solve is a question a Product Manager in coordination with the business stakeholders needs to establish. The Product Manager should perform initial customer interviews to understand their customer’s key pain points to validate the problem they are solving.

Dig Deeper into customer’s intentions

When getting customer feedback, it is essential to get to the root of the problem for the use case the customer has articulated. For example, the customer may ask can I export this data (predictive insight) into a CSV file?

One can think we need a CSV data export feature. Or maybe we can dig deeper, and the reason they may be wanting to do that would be they want to load the predicted insight directly into their CRM application for them to take action. So the real feature is for the predictive insight to drive some actions with some deeper integration with the CRM system and not exporting data into CSV. It is the job of the product manager to dig deeper into that insight.

Predictive Framework: Examples of predictive business problems

Framework for Product Managers to think about adding predictive insights by looking at some common/popular problems with good ROI.

The example / framework below are not models to build. They are examples of popular problems we can solve.

Simple example application

Say you have an existing product that you sell in the accounting space. Problems that you could solve that may differentiate your solution from the competitor would be problems like


  • Will this customer default on their payment?
  • Will this customer pay their invoice on time?


  • What would be the total billables next month?


  • Segmenting customers based on customer demographics & purchase behavior to better engage with them


  • Does this invoice look odd?

If you look at the above, we are not talking about what models to build or what algorithm to use or what data we need. We are identifying what problems to solve and which of these will be of value to your customers and will help drive your business.

Validate with your customers if the above questions add value to them. Validate if these help your business results. Once you got some good feedback, you can move on to the next step of putting a simple POC to validate the ideas for a product-market fit.


Start with the problem you are looking to solve. Define what benefits it has for the customer. Define how it helps drive your business results. Don’t be afraid to experiment. Experiments help in getting early feedback can help save time and money, and help correct the vision and direction of the product.