Empty-State Action Card on a Dashboard A/B Test

Support contacts asking “Where should I start?”, session recordings of users sitting on a blank screen and leaving, or accounts that never reach a first action after signing up all point to new users stalling on an empty dashboard. Users who see the card should then complete the predefined first action more often and reach their first meaningful data sooner. If the first action itself is hard or broken, fix that first; before launch, confirm the card can be dismissed and does not block access to any other feature.

First-action completion rate

Compare completion of the same predefined first action among eligible users assigned while their dashboard is empty.

Also track

• Time to first meaningful dataMeasure time from the empty state to the predefined first data event.
• Seven-day active useCompare all assigned users over the same seven-day activity window, not only those who follow the card.
• Support contactsTrack questions such as “Where should I start?”.
• Card dismissal rateUse dismissal as a diagnostic measure within the variant; there is no equivalent card in the control.

• CardAdd one predefined card that explains a practical first task.
• Keep constantKeep dashboard features, task requirements and other onboarding steps unchanged.
• Exit conditionRemove empty-state guidance once the relevant data or action exists.
• SegmentsApply the same eligibility rules when reviewing self-registered and invited users.
• Separate testsTest card copy, illustration or checklist-versus-single-action separately.

1. 1Compare first-action completion across all assigned eligible users, not only those who followed the card. If support contacts rise or seven-day active use drops, a lift in first actions is not enough to ship; read card dismissal only as a diagnostic inside the variant.
2. 2This card and this empty state are all the result speaks for. Review self-registered and invited users with the same eligibility rules, and compare seven-day active use over the same window in both arms to see whether an early gain lasts.
3. 3If only a limited number of new users reach the empty state, the difference may stay unclear for a long time. Take the users per variant that the sample size calculator gives for your current first-action completion rate and divide it by how many new users see the empty state in one arm each day. That gives you the run length up front; do not decide before it ends.
4. 4If the card uses sample content, label it clearly as an example, never present data the user has not provided as real, and do not show one segment's sample to another. Adding sample data also widens what the test changes, so plan it separately if you want the card's effect on its own.