Member-only story
Notes from Industry, BUSINESS SCIENCE
How (Not) to Fail At Your Data Science Project
The 4 design choices that can undermine ROI and impact
Over a year since the start of the Covid-19 pandemic, data scientists are still struggling to get their models back into shape. Every week or so, I see another article lamenting how the disruptions of the past year have negatively impacted machine learning models. Many organisations have stopped trying to adapt and are simply hoping to wait it out until we ‘get back to normal’.
They are going to be in for a shock when they finally realise that there’s no such thing as normal.
All of us working in data science need to recognise the failures that have caused these models to crash and approach our algorithmic design in new ways. We don’t have to keep making the same mistakes.
These 4 common project design choices in data science set you up perfectly for failure. Here’s your roadmap to model drift, minimal impact, and low ROI (or, even better, 4 things to avoid to set you up for success).
