The Data Science Lifecycle (CRISP-DM)
Part of the Data Science Roadmap
In short
Real projects don't go in a straight line from data to model. CRISP-DM is a map of the six stages, and of how often you loop back to fix earlier ones.
A widely used framework — business understanding, data understanding, data preparation, modeling, evaluation, deployment — describing how a real project actually moves (with lots of looping back).
Think of it like this
Cooking biryani for a party
First you ask who's coming and what they like. You check the pantry, chop and marinate, then cook. You taste it, maybe add salt and cook longer, and only then serve. Tasting and going back is normal, not a failure.
Watch it happen
A churn project through CRISP-DM
1/7Business understanding: the goal is to cut cancellations of a monthly plan by finding at-risk users early.
The big ideas
What you need to know
Start with the business goal
Write down what success means in business terms, like 'reduce cancellations by 5%', before touching data.
Preparation eats the most time
Cleaning, joining and building features usually takes far longer than training models.
Looping back is expected
Evaluation often sends you back to earlier stages. Plan your time for a few rounds, not one pass.
Deployment is part of it
A model that never leaves your laptop doesn't help anyone. Plan how it will be used from day one.
Save yourself the pain
Common mistakes
Oops
Judging the model only by accuracy in the Evaluate stage.
Do this instead
Check it against the business goal you wrote in stage one, like enough warning time or cost saved.
Oops
Budgeting one week for modelling and one day for data prep.
Do this instead
Flip it. Expect data understanding and preparation to take most of the project.
Your mission
Map a project on paper
Quick check
Your model scores well, but it predicts churn only one day before it happens. The business needs a week. What's the next step?
Go deeper
- 1Sketch a real or hypothetical project through all 6 CRISP-DM stages.
- 2Identify which stage usually eats the most time in practice (data preparation).
- 3Learn to budget your own project time expecting non-linear iteration, not a straight line.
More topics at this stop
What Data Scientists Actually Do
Data Scientist vs Data Analyst vs ML Engineer
Types of Data Science Roles (Research, Applied, Product)
Data-Driven vs Model-Driven Decisions
Common Industries for Data Scientists
A Day in the Life of a Data Scientist
The Data Science Career Ladder
Key Tools Overview (Python, SQL, ML, Stats)
Stuck on this topic? Ask an Insider
Get 1:1 guidance from people who've walked this exact path — on the InsideEdge app.
