At our company retreat last fall, I picked up a marker and walked to the whiteboard with a simple question: “What causes data science projects to fail?” In the ensuing ten minutes, I was soon looking for a larger whiteboard as my list grew rapidly. Everyone in the room was quick to throw out potential project killers – everything from internal politics to data quality.

We were able to have such a robust discussion that day because we all carried battle scars, and it wasn’t just academic knowledge we were sharing. For better or worse, data science, and all of its related tools and technologies, is as much experimentation as it is anything else – and most projects are fragile from the start. Over the years, we’ve compiled the various ways we’ve seen projects go awry, and we’ve actually built a pretty robust toolbox of mitigators and “de-blockers” that we deploy before a project ever begins. Consequently, when we debrief each year at our annual team retreat, we see a statistically significant difference in the projects we run through this rigorous process, vs. those where we jump right in.

Why would we jump right in knowing all this? Sometimes it’s just too tempting when a client approaches us with an urgent need, budget ready to go, and an excited business unit. However, time and time again, we have regretted not having the discipline to take them through this process that we know drives success.

“What are some of those failure points?” I’m often asked. I thought I’d share a few with you, and perhaps you have your own you’d like to add:

  • Lack of executive support
  • Quality of data
  • Availability of data
  • Reliability of data
  • Insufficient technology stack/resources to push a model to production
  • Overambitious models that require too much processing power
  • Legal concerns
  • HR concerns
  • Core team buy-in
  • Front line buy-in
  • Internal politics
  • Team capabilities
  • Recently upgraded systems
  • Systems in the process of being upgraded
  • And I would be remiss if I didn’t add coronavirus!

What did I miss? I’d love to hear your thoughts. Also, if you’re interested in seeing the tools we’ve developed to overcome these hurdles, we’re hosting a webinar this Thursday, April 30 at 1pm CDT. We’d love for you to join us!

 

DOWNLOAD OUR WHITEPAPER

"6 TIPS FOR BUILDING YOUR FIRST

DATA SCIENCE AND AI STRATEGY”

[[[[]],[[]],"and"]]
1

keyboard_arrow_leftPrevious
Nextkeyboard_arrow_right