When someone's first-choice domain is taken, which kind of alternative should we suggest?
Wrote the test brief covering design, user flow, metrics and success criteria
September to October 2026
Web analytics, spreadsheet
The question
On a domain-search site, one team was already testing two ways of showing alternatives when a searched domain was unavailable. A third approach, a top-positioned alternative already live on many partner registrar sites, had promising but indirect evidence. Should it get its own test, or join the one already running?
The method
I proposed adding it as a third arm to the running test instead of building a standalone one with its own budget, development cycle and timeline. The test randomly assigns users whenever their first choice is unavailable, and the top-positioned alternative always renders first, as it does on partner sites.
There is no untreated control arm, because adding one was out of incremental scope. So the test ranks three approved treatments. It does not prove that alternatives work at all. We read results as which arm performs best relative to the others, never as an absolute lift.
Before launch I wrote down the traps that could mislead the reading: consistent click-through denominators across arms, choice overload if arms show different numbers of alternatives, and relative versus absolute lift. The metrics are click-through on the alternative, click-to-purchase, time to conversion, bounce and cart abandonment. Search volume is context and the denominator.
The evidence
| Arm | Treatment |
|---|---|
| 1 | Exact-match alternative extensions shown |
| 2 | Variations of the search term on the default extension |
| 3 | Top-positioned partner-programme alternative shown first |
Two sources framed the decision. A consumer study found the placement roughly three times more impactful than a standard search-result placement on recall and influence, and one partner's data showed a large conversion lift against baseline. Both are directional. Neither was first-party or randomised, and this test fills that gap.
The result
The test is still running, so there is no ranking yet. The design is settled. It has three arms at incremental cost, and we fixed in advance which claims it can and cannot support.