Two warm cohorts, two different distributions of briefing rubric scores — and what each tells us.

§1 — What we looked at

Two cohorts, defined precisely from Salesforce.

We isolated the prospects the CEO asked about and looked at briefing quality vs. outcome for each cohort independently. Two Salesforce fields define them:

Notable: zero overlap — no opportunity is both a Referral source AND an Alumni rejoin. They are cleanly separable cohorts. Each briefing was scored 0–2 on each of 8 sections (Cheat Sheet, How Set, Past Calls, Existing Members, Why Them, Triggers, Group-Specific, Alumni-Specific) for a total possible score of 16. Scoring done by an LLM-based agent, blind to outcome.

Headline cohort win rates (no rubric cut): Referrals win 55% (n=53, avg rubric 9.5); Alumni rejoins win 60% (n=25, avg rubric 11.5). The cold/Sales-Process baseline is 26% (n=163, avg rubric 9.0). Referrals close at 2.1× the cold rate; alumni at 2.3×. The question this analysis examines is how rubric scores associate with outcomes within each cohort — and the data we can observe is different for each.

How outcomes are determined: Each opportunity's outcome is its Salesforce IsWon flag — the canonical revenue truth. This sometimes diverges from the Drive folder a briefing was filed in (e.g., a deal initially treated as won, then canceled mid-term, ends up SF-lost despite Drive-folder placement). We use SF as ground truth on every record.

§2 — The Referral cohort

Rubric scores and win rates show a 21-point gap between briefings scoring ≥ 8 and those scoring < 8.

§3 — The Alumni rejoin cohort

Rubric scores cluster at the high end of the scale; the below-threshold cell is too small for a meaningful comparison.

§4 — Why the two cohorts give us different data

Two structural differences in the data we have for each cohort.

Two observations about the dataset itself:

1. Rubric-score variance differs sharply between the two cohorts. Referral briefings span the full rubric range (4 score 0–5, 14 score 6–8, 23 score 9–11, 12 score 12+; avg 9.5). Alumni briefings cluster at the high end (2 score 0–5, 1 scores 6–8, 8 score 9–11, 14 score 12+; avg 11.5). To observe a relationship between rubric scores and outcomes within a cohort, we need variance in the rubric scores. The Referral cohort has it; the Alumni cohort does not.
2. Salesforce data structure differs between the two cohorts. A referral arrives with a named recommender — that recommender appears in the briefing as a champion. An alumni rejoin arrives with prior call records and a prior decline reason — that history appears in the briefing as past-call context and alumni-specific notes. Two different starting positions for briefing prep, which may explain why the rubric-score distributions look different.

What this means for analysis: rubric-vs-outcome comparisons within the Referral cohort have enough variance to produce a defensible result. The same comparison within the Alumni cohort doesn't, because the inputs are too narrow.

§4.5 — How rubric-vs-outcome differs between Referral and Cold cohorts

Same rubric, two cohorts, different patterns.

With n=163 cold deals scored on the same 8-section rubric, we can compare the rubric-vs-outcome relationship across cohort types directly.

Two patterns in the cross-cohort comparison:

1. The within-cohort rubric-vs-outcome relationship is steeper for Referrals than for Cold. Within the Referral cohort, the gap from rubric < 8 (40%) to rubric ≥ 8 (61%) is 21 pts. Within the Cold cohort, the same cut produces 22% vs 28% — a 6-point gap. The rubric scores associate more strongly with outcomes in the Referral cohort than in the Cold cohort.
2. Within the Cold cohort, win rate is roughly flat across the 6–8, 9–11 buckets. 6–8 wins 27%, 9–11 wins 26%. Only the 0–5 bucket (6% win rate) and the 12+ bucket (38% win rate) sit notably outside that range. The mid-to-high section of the rubric doesn't differentiate cold-cohort outcomes the way it does for referrals.

What this comparison shows: the 21-point gap observable in the Referral cohort isn't reproduced at the same magnitude in the Cold cohort. The relationship between rubric scores and outcomes is stronger in the Referral data than in the Cold data.

§5 — What the data supports examining next

Three areas where the dataset suggests further work would be productive.

1. Referral cohort, rubric ≥ 8 cut. The data shows a 21-point win-rate gap between Referral briefings scoring ≥ 8 (61%) and those scoring < 8 (40%), on cell sizes of 38 and 15 respectively. If 8 is used as a workflow checkpoint and the 15 below-threshold briefings closed at the above-threshold rate instead, the cohort would have produced approximately 3 additional wins.
2. Alumni cohort, original-decline pattern. Within-cohort rubric variance is too narrow to support the same comparison method. The dataset suggests a different question would be more productive: among alumni who re-engaged, what original decline reasons (prior Won_Lost_Reason__c), time-since-decline values, and role-change patterns associate with re-engagement outcomes? That's a separate analytical pass on a different field set.
3. Referral cohort, section-level analysis of below-threshold briefings. The 15 referral briefings scoring below 8 may share specific section-level patterns (e.g., S3 past-call history empty, S4 existing-members count low). A breakdown by section would identify where the rubric points are concentrated. Comparing those patterns against the Alumni cohort's section-level distribution, where rubric scores are uniformly higher, could surface what differs between the two briefing-prep paths.

§6 — What we're not yet confident in

Honest list of where the data isn't yet strong enough to commit.

• Referral cohort, below-threshold cell at n=15. Real but not large. If the next 15 below-threshold briefings closed at 60%, the gap would compress to single digits. The direction is robust; the specific 21-point magnitude is sensitive to additional sample.
• Alumni 9–11 vs 12+ inversion on n=8 / n=14. Both cells small. The two candidate explanations (rubric scores reflecting deal difficulty, or rubric scores affecting outcomes) can't be distinguished with rubric-only data at this sample size.
• The Referral source classification mixes five sub-types. Member Referral dominates the cohort; Advising Member Referral, Advisor Referral, Alumni Referral, and Prospect/Suspect Referral are tiny. The cohort behavior is driven almost entirely by Member Referral. A sub-type breakdown would refine the cohort definition but cells get small fast at this n.
• The analysis associates rubric scores with outcomes; it doesn't establish causation. Higher rubric scores may reflect easier-to-close deals receiving more populated briefings, or higher-rubric briefings causing higher close rates, or both. The cross-cohort comparison (§4.5) is useful here: if reverse causation alone drove the result, we'd expect the same pattern in the Cold cohort; we don't see it at the same magnitude. That weakens but doesn't eliminate the reverse-causation candidate.
• Recovery-pass methodology. The cohort grew from 67 → 78 briefings via a relaxed-match recovery against the full Salesforce dataset (single-lastname relaxation, first-name disambiguation, broader stage filter). 11 of the 12 recoveries that affect the cohort came in with SF IsWon=false while sitting in Drive's "won" subfolder (often because of mid-term cancellations). We use SF IsWon as the canonical outcome throughout.

§ Citations

Direct links to all briefings in each cohort.

Each row links to the briefing's source document on Drive (opens in a new tab). Sorted by outcome (won first), then by rubric score (highest first). Use this to verify cohort membership, audit individual briefings, or sample for follow-up review.