At Blue Yonder we are about to introduce a new data-driven product called Luminate Cognitive Inventory. Leading the data-science team’s contribution to this endeavor, I would like to share some of our experiences gained during the journey from ideation to go-live. Using this real-world example, I will walk you through the various technical and organisational challenges we have faced. Starting with a business problem statement, we must navigate and balance the demands of data wrangling, model building, a customer facing proof-of-concept, and finally a production implementation. To make you curious, let me just mention one piece of advice we would give before you embark on a similar journey yourself: Don’t try to build a product without a first real customer. But if you have such a customer, it is very easy to slip into the trap of building a customer project instead of a product. Balancing this thin line will have much bigger impact on the success of the product than some few percent of model accuracy improvement.