The Report That Counted Twice
You built a Revenue Explorer program performance report for a webinar. The program had 2,000 members. The report shows 2,400 'Responded' and 2,100 'Attended.' You know only 1,800 people attended. The program has: a registration form (fills a 'Registered' status), an attendance import that sets 'Attended,' and a follow-up email with a click that sets 'Engaged.' You also have a smart campaign that changes status to 'Attended' when a person's behavior score exceeds 50 during the webinar week. Your report uses 'Program Status' as the success metric and 'Responded' as any status beyond Registered. A colleague says the report is 'just wrong' and wants to rebuild it from scratch.
1. Explain why the report shows more Responded and Attended than reality, using the program mechanics.
Read the worked answer
The report is counting status changes and inferred attendance, not verified attendance.
'Attended' overcount (2,100 vs 1,800): Two mechanisms set 'Attended.' The attendance import sets it for actual attendees, but the behavior-score smart campaign also sets 'Attended' for anyone whose score exceeded 50 during webinar week - which includes registrants who never joined but were active elsewhere. That inflates Attended by ~300.
'Responded' overcount (2,400 vs 2,000 members): 'Responded' is defined as any status beyond Registered, which includes Attended and Engaged. A person can move Registered → Attended → Engaged, and each transition is a status change. If the report counts status-change events rather than unique people, multi-step progressions are counted multiple times. Also, 2,400 > 2,000 members means the report is almost certainly counting events, not unique members, or including people who were removed/re-added.
Conclusion: The report is not 'broken' - it is faithfully reporting a flawed status model and an event-vs-person counting choice.
2. How would you fix the measurement so the report reflects true attendance and response, without rebuilding everything?
Read the worked answer
Fix the model and the metric definition, not the whole report.
Separate verified attendance from inferred. Stop using behavior score to set 'Attended.' Create a distinct status 'Inferred Interest' or leave score-based leads at 'Registered' and tag them with a field. Only the attendance import should set 'Attended.'
Make status progression non-destructive. Define a clear status ladder (Registered → Attended → Engaged) and ensure each person holds exactly one current status. The report should count unique people currently at or beyond a status, not the number of status-change events.
Redefine 'Responded' as unique members whose current status is Attended or Engaged (or who have a success flag), not 'any status change.'
Use a success flag (e.g.,
Webinar Success = True) set once, so success is a boolean per person and cannot double-count.Validate: After fixes, the report should show Attended ≈ 1,800 and Responded ≤ 2,000. If it still exceeds membership, investigate removed/re-added members and report filters.
Why not rebuild: The data model is the problem; a new report on the same statuses would reproduce the same errors.