Machine-sourced against human-sourced, compared on the only thing that settles it — closed-won opportunities and revenue. Name-use permission requests are excluded throughout: they never reach a member and have no win state, so counting them would flatter every rate.
Why there is no “we asked the member”
column. An earlier version of this page split human sourcing using the Data Source field. That split was
invalid. The field is not fixed at record creation — the “Direct Member Referral”
share climbs from 10.5% of records at Research Submitted to 64.0% at a win, and peaks at
69.0% on No Response From Member, which is a loss. It tracks how far a record travelled,
not where it came from, so outcomes compared across it were circular. Selma is safe to split on
because it is machine-assigned at creation.
| Channel | Records | GD decided | GD approval | Referrals made | Made rate | Follow-through | Opps | Closed | Won | Close rate | Revenue | Wins / 100 rec | Revenue / record |
|---|
| Measure | Count | Read |
|---|
Shown for transparency only. These labels drift with funnel stage, so nothing on this page computes an outcome from them.
| Data source | Records | Share of research channel |
|---|
Sorted by revenue per sourcing record. Click a row to open that group.
| Group | Group Director | Records | Selma share | Human-sourced won | Selma won | Close rate | Revenue | Revenue / record |
|---|
Generated from referrals_1_25.csv joined to Salesforce opportunity
outcomes on the referral item number. Rebuild with
analyses/group-drilldown/build.py.