Combining booking and application into one experience increased the reported complete-application rate from 14% to 17.5% across 1,409 participants over five days. Those rounded rates imply a 25% relative increase, while the historical article says 26%; exact counts are needed before restoring a precise lift claim.
The before-and-after journey
| Stage | Control | Combined treatment |
|---|---|---|
| Choose a call time | OnceHub booking form | OnceHub booking form |
| Provide application details | Separate mandatory Wufoo form after booking | Questions integrated into the booking flow |
| Repeated personal information | Asked across two systems | Intended to be collected once |
| Incomplete application rule | Call could be cancelled after two hours | Handled in the combined flow |
This was a package change: one fewer handoff, removal of duplicated personal fields, movement of application questions into OnceHub, and different failure/cancellation behavior. The experiment can estimate that package’s observed effect. It cannot prove that the number of steps alone caused it.
Observed result and arithmetic
The current record reports 14% completion for control and 17.5% for treatment. The absolute difference is 3.5 percentage points. Dividing that difference by the 14% baseline gives 25% relative lift from the rounded values. A reported 26% may come from unrounded rates, but per-arm counts and precise output are absent.
The article also says 21% of control users selected a time and abandoned the second step, then later describes “a third” dropping before completion. These may use different denominators, but no denominator is shown. This revision does not repeat either as a settled abandonment rate.
What a reliable implementation must preserve
Create a field map before merging systems: identify the source of truth for contact details, qualification answers, consent, selected time, timezone and routing fields. Remove true duplicates, not information required for scheduling, compliance or qualification. Confirm that validation errors retain the chosen slot and entered answers.
Test expired slots, double submissions, API failures, incomplete required fields, cancellation timing and CRM delivery. Record a completed application only after all required data is accepted. A calendar selection is a funnel start, not the final conversion.
When separate steps can still be appropriate
A separate application may be useful when people need to save progress, gather information, receive conditional questions, or complete qualification before scarce calendar inventory is reserved. The design goal is not universally “one form”; it is a coherent handoff with no unnecessary repetition and an outcome definition aligned to the business decision.
Evidence needed for the full case study
Recover arm assignments, complete applications, slot selections and qualified or attended calls using a common assigned-user denominator. Reconcile 25% versus 26% and the two abandonment descriptions. If purchases or sales quality were not measured, do not infer them from form completion.
The urgency-question case asks whether a qualification field adds useful information; it should not be used as proof that every removed field is beneficial. Use the A/B testing metrics guide to define the complete and downstream outcomes before a replication.
