Too short a run
One week does not cover a booking rhythm. Two full weeks are the minimum, often four are needed.
TRFor travel and booking
Season beats
every variation.
On booking sites everything fluctuates: weekday, holidays, weather, competitors’ prices. Ignore that in your test design and you measure the season instead of the change.
The pattern
Three things that distort
the result.
Campaigns in the middle of the test
A discount period changes the visitor mix, visible in the segments by source.
Only the booking measured
Many visitors compare first and book later. Entering the booking flow is the more honest primary goal.
Concretely
Where to start
Simplify the search
Fewer fields, dates pre-filled. Goal: search started.
Price with the total
Total instead of per-night price. Goal: booking started, guardrail: checkout abandonment.
Show availability
X left, honestly and verifiably. Goal: booking.
Move cancellation terms up
Flexibility as a selling point. Goal: booking.
Measuring
The right goal
The primary goal is entering the booking flow, or the completed booking with revenue, depending on which happens often enough in your test window. At the end, check the segments by source: if the gain comes only from one campaign, it is not a gain.
Sample size and runtimeMatching platforms
Other industries
Elsewhere it looks different.
Make your next change an informed one
Less guesswork.
More now we know.
One question is a good place to start.