Introduced on day 1
Baseline
The strongest simple method a careful analyst would actually use, computed with only the information available at decision time. A model is worth adopting only if it beats the baseline on unseen data by more than its cost.
In detail
For per-flight no-shows a good baseline is the average for the same route and weekday, computed from past flights only. For daily passengers with a weekly pattern it is the seasonal-naive forecast: the same weekday last week.
A baseline must be strong, simple, auditable and leak-free. "Predict zero" is too weak to prove anything. A route average that includes the evaluation period is leaky: it has seen the answers you are about to score.
The baseline also sets the business case. If a gradient-boosting model reduces error only slightly versus the lookup table, its build, serving and monitoring cost can exceed the value of the gain. Report the comparison, not the model's error alone.