Prior to our interview, I try to imagine that I am about to speak to one of my close friends who happens to be a theoretical physicist; bracing myself for a conversation, where I as a humanities major will have to ask a lot of rudimentary questions.
But my conversation with data scientist Kristen Kehrer from Data Moves Me on data and the state of actionable machine learning in organizations today, doesn’t turn out that way. Far from it. In fact, the notion of data scientists as nerdy builders sitting all quietly in the corner, is what Kristen is changing through her work:
I see myself as a bridge between the models that I build and the business. There is a huge problem where data scientists are seen as data wranglers and not communicators. They sit in the corner and build a model and then have difficulty explaining the implication or how this is actionable in a clear way to the business. To me, the goal is to bridge the gap between the amazing things we are now able to do with data, and what organizations can integrate and use within the business; to communicate how data can empower strategy and deliver value to customers.
The state of data and machine learning in organizations today
Organizations’ interest in data and machine learning has been rapidly growing. According to the study The State of Machine Learning Adoption in the Enterprise published by O’reilly in 2018, 49% of organizations report that they are exploring or “just looking” into deploying machine learning, while a slight majority of 51% claimed to be early adopters (36%) or sophisticated users (15%).
However, interest does clearly not equal adoption, as thought leaders on machine learning Megan Beck and Barry Libert argues in Forbes:
Despite a great deal of lip service and a small amount of capital invested, most corporations are still not data-driven, nor do they use machine learning (ML) and artificial intelligence (AI) to guide their strategic investments in business models.
Kristen wouldn’t go as far as calling the interest shown from organizations lip service, but she agrees that there is still a gap between possibilities within the field, and how far businesses have come in terms of being able to put these to use. It’s exciting, but for many organizations really investing full on in data and machine learning can seem like space exploration, in which the immediate business value can also seem somewhat blurry:
Business and Organizations today usually have access to a lot of data, so first it can be difficult for them to determine what is valuable, where to focus, and how they can make it actionable. A big part of this comes down to clear communication. Not just building machine learning models but building trust.
Becoming a thought partner
So, organizations can be overwhelmed by the possibilities offered by data and machine learning. With this point being at the top of her mind, there are things that Kristen would have done differently, looking back at some of her projects.
For instance, showing the functions underlying the model to stakeholders simply causes confusion. Not only because “the techy part” can be difficult to understand, but more so because the functions do not really show the greater picture of business value.
Kristen’s job becomes about showing customers how machine learning solutions can help them achieve their goals. Or inform them, so they can set new and better goals. It becomes a process of continuously asking why, and not focusing too much on the what:
A good data scientist should be a thought partner, and not just a machine learning developer. We must build the relationship with stakeholders, and clearly communicate the value to their jobs, so they become advocates of your work. Getting involved early in the discussions about what the stakeholders want to achieve and having full business context is also a big part of helping this relationship develop.
This work should always be seen as a long-term investment. Continuously analyzing data, seeing opportunities for optimization and how it can benefit the business. Potential is fine, but Kristen stresses the need for focusing on how to make actionable solutions here and now:
The data is valuable to my work, but what is valuable to businesses and organizations, is how that data can be leveraged to shape their strategy, make the business more efficient and deliver ROI. This means that my job is never simply to show people the data, but to communicate the story it tells.
If you are interested in a deeper understanding of how to make data actionable and use it to create business value, consider joining Kristen for the Boye 19 Brooklyn conference in May, where she will be giving her keynote titled Models are Not Black Boxes on bridging the communication gap between data science and the business.
Additional reading from Forbes:
The State of Machine Learning (Sep, 2018)