Choose the question before the date filter
Bookings created this month answer a different question from stays arriving this month. For a booking cohort, count eligible bookings created in the period and identify which later cancelled. For an arrival cohort, follow the reservations due to arrive in the period, including those cancelled before arrival. Label the date basis and avoid comparing unlike cohorts.
Keep the denominator visible
Show cancelled bookings alongside the total eligible bookings, not just a percentage. A change from one cancellation to two can look dramatic when the sample is small. Decide how to treat inquiries, test reservations, modifications and missing dates, and document that choice. Keep unknown records visible as an exception count instead of silently assigning them to a month.
Segment before changing the policy
Compare channels, booking lead time, stay length and rate conditions using the same definition. Investigate whether a channel has a different cancellation window or guest mix. A cancellation rate alone cannot prove a policy caused the change. Also consider whether cancelled nights were resold, how close to arrival the cancellation happened and whether any collected amount was refundable.
Turn the review into a bounded action
Review a small set of underlying reservations before changing a policy. Record the proposed change, the owner and the comparison period. Recheck after enough new bookings have accumulated, keeping the original definition. Do not treat a lower cancellation rate as a success if it came with fewer bookings or worse net revenue. Operational reports are a starting point for investigation, not an automatic instruction to tighten terms.
Practical operations guidance, not legal or tax advice.