Procurement’s not just about cutting deals anymore – it’s about…
Artificial intelligence (AI) is changing the way businesses operate, especially in the supply chain sector. Nathan Cunningham, Founder and CEO of WIPP Data, has seen firsthand how AI and machine learning can revolutionize business development. With a background in supply chain from Tesla and now focusing on fractional data science, Nathan brings a unique perspective on the power of AI. Nathan shares his journey, a recent machine learning success story, and advice for others looking to implement AI-driven solutions in their own organizations.
You recently worked on a machine learning forecasting solution for one of your clients. Can you tell us more about that project?
When you first launch your own machine learning (ML) project, you're often surprised by what it can do. I had some experience working with machine learning at Tesla with moderate success, but this recent project far exceeded my expectations. The client had 10 years of potential customer data, and they wanted to predict a specific behavior in that dataset. We were able to narrow in on certain companies and create a model that predicted how likely it was that a specific event would switch from a no to a yes. It was a huge amount of data, but the success we found was mind-blowing.
What business challenges was the solution addressing?
The client had a massive dataset with over 10 years of potential customer information. There was a specific column they wanted to predict – essentially whether a certain event would change from a no to a yes. It was a classic predictive problem, and while we had to deal with a lot of data, the real challenge was narrowing down the right information without overwhelming the model. Finding that balance was key to the success of the project.
What were some of the key success and learning points throughout this project?
One of the biggest realizations was that machine learning isn’t just about feeding a ton of data into a model and expecting it to work. There’s a sweet spot where you have enough data for the model to learn, but not so much that it just memorizes the dataset. For example, one of the challenges we faced was dealing with data columns with large ranges, like total assets. If you give the model too wide a range, it won’t know how to process it effectively. We had to aggregate that data to optimize the model and, once we found that balance, it really started to work well.
Did the client already know which data would be good predictors, or did you figure that out during the project?
The client had been in business for a long time, so they had a gut feeling about which data points would be relevant, but we were both surprised by the results. They thought certain variables would be the most important, but when we ran the model, those factors ended up ranking lower than expected. It was a good reminder that even experienced professionals can have blind spots when it comes to data, and AI can help uncover unexpected insights.
How do you see this machine learning model evolving over time with the client?
I think there’s definitely room for improvement, especially by incorporating additional data sources. Right now, we’re only pulling from one data source, but as the model continues to deliver positive results, we can expand and enhance it. There’s also a lot of potential for post-prediction filtering. For instance, even if a company has a high probability in the model, the client might not want to pursue them for other reasons, like their location. We can refine the model further to account for factors like that and continue improving the pipeline.
What advice would you give to others looking to implement a machine learning solution for business development?
My biggest piece of advice is don’t do it alone. I tried to tackle a lot of it on my own and hit a brick wall. Once I brought in a few more people, the project really took off. Even if it’s just bringing someone in for advice, having another perspective is invaluable. Also, make sure you have a well-defined use case before you dive in. AI is inherently imperfect, so you need to know exactly what you want to achieve. In business development, precision isn’t always necessary – you need trends and insights more than perfect answers, which makes AI a great fit.
Nathan Cunningham's journey with machine learning in supply chain data is a testament to the power of AI when applied strategically. His success highlights the importance of collaboration and defining clear use cases for AI projects.