As a data scientist, have you ever been frustrated that your stakeholders don’t see the value that you bring to the table? You may ask yourself, “How far should I go in explaining the work I do or what my models are doing?” If that sounds like you, then pay close attention to this post and the next, as they are all about improving collaboration between data scientists and other stakeholders.
This is a two-part post: This article covers underlying assumptions and gaps in understanding that cause friction between data scientists and stakeholders; my other article, offers concrete steps for better collaboration.
Machine Learning (ML) models are inherently complex and hard to explain. Data scientists know that the events between data input and output generation are not easily mapped to an explainable process. Before the adoption of ML techniques, the world was much simpler; it was easier to explain how the output was generated. Back then, data science operated on a rules-based system; in fact, it still does, even though the rules are now more complex.
More importantly, in the world before ML models, the rules were governed by stakeholders throughout the process. (Note that my usage of the word stakeholder is broad...this could be a general manager, business owner, marketing lead, product manager, etc.) While this is no longer the case, stakeholders can still have a lot of say in the ML process. For example, stakeholders still hold the key to the input data in many environments. Sometimes they even own much of the process that leads to data output.
Further, a lack of understanding can lead to miscommunication, which damages trust between the parties. This lack of trust can seriously impede a data scientist’s ability to provide the right level of support throughout the end-to-end process of model building (i.e., collecting, building, deploying, and iterating on the models).
Oftentimes, management support for resources/time/capital allocation is needed to get better outcomes from ML models. Without such investments, the results of ML models are either subpar, or even worse, a waste of time. Remember, a model that predicts with 50% accuracy is of no use.
The above are some of the primary gaps from the stakeholders’ point of view (POV) resulting in misaligned expectations and a lack of trust. While these gaps impact all parties involved, here are the implications from a data scientist’s point of view.
Lack of support and guidance:
I have heard data scientists express that when the company started setting up a data science team or hiring new data scientists, there was so much excitement; however, the support and enthusiasm they experienced at the beginning faltered after a few months or quarters. Of course, not to say all companies are alike, but this pain is most often experienced when companies start the journey towards incorporating data science principles/techniques in their products/processes. Again, if the stakeholders feel they are not getting what they want due to communication issues, data scientists will end up feeling let down and/or neglected.
Companies often hire the wrong type of data scientists or the wrong level of seniority for a project. This usually happens when the company is getting started with data science and has no clear understanding of what it wants from this team/role/person. This misalignment ends up further souring relationships while wasting time and effort across the board.
The set of gaps listed above are indeed fixable. It takes time, effort, learning, and setting up some frameworks so that both entities - data scientists and stakeholders - can foster better collaboration and achieve more together. Check out some proposed solutions for fixing communication gaps between data scientists and stakeholders in my next article!