Based on domain experience, ask specific and meaningful questions.
Define all terms precisely within the context.
Is the organization even collecting the right data?
Do I even have access to the relevant data for the problem?
Can I get the stored data to where the modeling takes place?
Get all the data in a single data structure (typically a data frame).
Understand, organize, and munge.
Identify features and targets.
Apply Statistics, Machine Learning, and Deep Learning algorithms.
The goal is to add value. Is the value in a report, presentation, or code?
How are the models going to users? Going into production? Transfer code to software development team? PMML? Create API?
What does the system update when the data changes?