Recently Pariveda hosted the Data Science Innovation Summit where a long-standing hypothesis was once again confirmed. The Data Science Product Owner continues to be a key missing link in modern analytics projects, including Artificial Intelligence (AI), Machine Learning (ML), Predictive Analytics (PA), and others.
Here is why:
- Projects fail because they are not sufficiently linked to tangible and well-understood business value. Analytics project teams must understand the business value and work as the intermediary between the Data Science team(s) and the Business Stakeholders. Analytics project teams must be able to transform business value into model designs for the Data Science team(s) and must be able to interpret and translate model constraints for the Business Stakeholders to allow them to pivot during a Test and Learn cycle. This is often more than the Data Scientist is prepared for and models are left to die in the “build it and they will come” pasture.
- Projects fail because they cannot sufficiently influence or improve upstream process and systems. The business value of advanced analytics projects is limited by the availability of quality, relevant data to input into decision-making model(s). First-revision models often uncover upstream data that needs to be cleansed, aligned, patched and “wrangled” (as Data Scientists often complain about). First-revision models also typically expose business processes and technology adaptations that are needed to feed the model to make it valuable. For example, a pricing model may need a new field in CRM to allow appropriate segmentation of the data or a data point may need to be added to records in one system to allow it to be linked to data from another system. Again, the Data Scientist is often not able to influence the needed change and models are left to die in the “this is harder than we thought” pasture.
The Data Science Product Owner is a cross-functional role that requires many “generalist” traits that are uncommon but critical to the success of an advanced analytics project. There are elements of a “storyteller” role within the Data Science Product Owner role. Without recognizing the risk points above, the Data Science team or the Data Science team lead are responsible by default, without the specific expectations of being a Product Owner. Drawing on Agile methodology, the Product Owner is typically a project’s key stakeholder. Part of the Product Owner responsibilities is to have a vision of what he or she wishes to build and convey that vision to the scrum team.
“ The best way to predict the future is to create it.” – Peter Drucker
Of course, these aren’t the only reasons analytics projects fail, but they are the big ones. There is buzz in the market and participants of last week’s Data Science Innovation Summit agree and the topic was mentioned by each conference speaker, raised in audience questions and follow-up social media conversations. This timely debate confirms there is still a challenge for companies wanting to invest in advanced analytics.
How will you define the Data Science Product Owner for your next advanced analytics project?