Dr. Wood's Rules for Successful Projects

Understand the use case. Not just the stated requirements. Know why you are doing the research/analytics and what the exact problem is to be solved. Only when you know what and why can you determine appropriately how to perform the research/analysis.

Know what is essential, what is good to have, and what is nice to have. Just because something is interesting does not make it a good project.

Understand how the end-user will use the results. Make the deliverable(s) for the results not only useable but user-friendly and intuitive.

Learn to know when it is "good enough". There comes a point of diminishing returns.

Learn how to make the end product the most useable (see point 3).

Avoid noise. If something is distracting rather than helpful, remove it or move it.

Avoid perfection. There is no such thing.

Spend your time and resources solving known problems, not finding problems for your solutions (be decision-driven, not data-driven or  methods-driven)

Test, test, test. Test it until it breaks. You need to know when it will go wrong, not just when it works.

Use the fail-fast approach. Not all projects can be successful, so use an approach that identifies projects that will fail quickly

Use your own products. There is no better way to understand the pains you cause the end users.

Know what you do not know. And don't pretend you know something, or have solved something, when you haven't.

There are things you know, things you don't know, and things you don't know that you don't know. The things you don't know that you don't know are the most dangerous. Always be on the lookout for these and for unintended consequences.

Don't work in a silo. No one is perfect or knows everything. It may take longer to work with others (particularly diverse teams, with diverse backgrounds and ideas), but the end results will be higher quality and less error-prone.

Always go back to the fundamentals. Don't rush to use your favorite methods or the latest and greatest algorithm. Fundamentals can often solve a problem faster and more robustly than using advanced or complicated methods.