“There are no facts inside the building so get the heck outside.”
Steve Blank, the father of the lean startup methodology, built his career on the idea that the best products don’t come from product managers or designers, but from customers. No matter how good of a product mind you are, he argued, you have to talk to customers to be successful.
Data scientists and machine learning engineers - and especially those working on industry-specific problems - should follow the same advice. As data scientists, we take pride in our craft, and often gravitate towards the things that made us interested in data in the first place: mathematical models, engineering challenges, statistics, data visualization, and other tools of the trade.
But by turning inward and focusing on our areas of expertise, we miss what’s in our blindspots: the context around the problems we’re trying to solve. After all, no model or prediction is based on data alone. Our work is also based on dozens of assumptions we make - about which features are important, on casual relationships between features, or on how to measure the accuracy or success of a model. To make good assumptions, we can’t just be experts in data science; we also need to be experts in our industry.
Even more importantly, the more we understand the business problem, the better we are at identifying the areas where our skills can be best applied. As Kathryn Hume put in the Harvard Business Review, when trying to find machine learning opportunities, “machine learning scientists can’t work in a vacuum; business stakeholders should help them identify problems worth solving...” The more we know about these stakeholders and the challenges they face, the more fruitful the collaboration, and the more valuable our work.
So leave your “office” and talk to your customers (students, professors, administrators, etc). Learn about the people using your products. Figure out what motivates them and what frustrates them. Don’t try to solve hard business problems with data science - make data science easy by understanding the business problem.
Mode believes data science and machine learning teams should empower everyone in their organization to make better decisions. We deliver to thousands of teams daily with a streamlined end-to-end workflow, built the way data experts work.
"It used to take days for our analysts to go from a request for data to a report. With Mode it often takes minutes."Denis Zgonjanin, Engineering Lead — Data Presentation, Discovery, and Governance
Shopify augments its data stack
See how Mode enhances day-to-day productivity and complex, ad hoc projects for thousands of users.Learn more
Molecule makes an impact
Molecule uses Mode White-Label Embeds to deliver real-time data to
Mode plays nicely with others so you can deploy the best possible solution for each aspect of your infrastructure
Refresh data anytime
Help everyone explore data
Deliver data right where you need it
Get the right data to the right people
Live in-product chat