AI Glossary Term
What does “Baseline Model” mean?
A simple reference model used to compare whether more complex approaches actually improve results.
How Baseline Model works
Example
A team predicting customer churn first builds a rule: flag anyone who has not logged in for 30 days. Their new neural network must beat that simple rule by a meaningful margin to justify its extra cost and complexity.
Why it matters
Without a baseline you cannot tell whether a sophisticated model adds value. Simple baselines often turn out to be surprisingly hard to beat.
Common questions
What makes a good baseline?
Something simple and cheap, such as predicting the majority class, using the previous value, or applying a basic rule or linear model.
Do I need a baseline for every project?
It is strongly recommended. Reporting a model's score without any reference point makes the result hard to interpret.
What if the baseline wins?
Then the extra complexity is not paying off, and the simpler approach may be the better choice.