When Your Data Science Project Fails

Why Projects Fail

• No data. Make sure you've thoroughly looked into the data you have available, and bake in an early “go/no go” step in the project to review project feasibility with stakeholders.
• No signal in the noise. It's usually more surprising for there to be a signal than for there not to be a signal.
  • You can try to reframe the problem or change the data source to salvage the project. If the simplest method cannot detect any signal, the more complex ones won’t be able to either.
• The customer didn't want it. Reduce this by working closely with stakeholders. Deliver a minimally viable product (MVP) and iterate on it.

Managing Risk

Risk can come from the individual project and the combination of projects you're working on at one time. Work on multiple projects at once and diversify your portfolio (like stocks - a lot of low-risk and some high-risk). Bake in early stopping points into projects - for example, the project can be scoped so that if, after a month of searching, good data can’t be found, it’s considered to be infeasible and scuttled. If the expectation that it might not work out is presented early, ending the project is less surprising and less costly.
Plenty of valuable data science contributions at companies have come from people trying something new, and if as a data scientist you try to avoid projects that could fail, you’re also avoiding potentially big successes.

What to do When Projects Fail

Document Lessons Learned

You can learn from every failure.
Ask yourself (and the team):

• Why did it fail?
• What could have been done to prevent the failure?
• What did you learn about the data and the problem?
  See if you can produce any actionable takeaways to make structural changes to improve results over time.

Consider Pivoting the Project

Can you repurpose what you've done so far? This requires a lot of communication with stakeholders (more than usual). You're back at square one and re-scoping the project.

End the Project (Cut and Run)

Can't pivot? Kill it. Don't get sucked in by the sunk-cost fallacy.

Communicate With Stakeholders

Increase communication as the project fails. Don't give your stakeholders any surprises. Being honest builds trust and relationships.

Handling Negative Emotions

Give yourself grace. You're on a treasure hunt, not building houses.

Sources: 1