Build an A/B/C (3-condition) test to test:
- The default experience ("test-group-a")
- The search box from the first A/B test ("test-group-b")
- The search box from the first test, plus typeahead improvements ("test-group-c")
The logic for admitting people to the test:
- 1 in 200 people get into the overall schema
- Of these, 1 in 10 get into the test. The remaining 9/10 get the default experience and are not in a test group - they are for dashboarding
- Of the 1 in 10, 1/3rd of users are distributed to each test group.
We should keep the same client technical requirements that the default eventlogging has, unless doing so would damage the quality of the data. In that case, we should keep those same technical requirements for the non-tested (9/10) group. If that's not possible, let me know.