How experiment management can improve the ROI of your machine learning projects

When a data engineer is building a pipeline she can usually estimate how much time implementing a step in that pipeline would take. To a large extent, that information is available to her. When a data scientist is doing data exploration, the information needed to estimate the time to extract actionable insights is hidden in the very data that she is exploring. It is not available prior to the exploration. Not to mention, discussions with other stakeholders both technical and business side that arise because of the analysis which makes it even harder to estimate.