Group-Level Cut · Output of group_cut.py

How does briefing-craft signal vary by community?

A breakdown of the 184-briefing rubric scoring by W50 community group, plus the thick-company effect (existing-member density at the prospect's company vs. win rate).

184briefings scored

85 / 99won / lost

19distinct W50 groups

5groups with both W and L ≥ 4

§1 — Per-group readout

Section deltas vary substantially by community.

Read this as: which W50 group does the briefing belong to, how many of its briefings won/lost in our sample, the total rubric-score delta (out of 16), and which single section showed the largest Won−Lost gap.

Group	W	L	Total Δ	Top single-section delta
FR50	3	5	+5.07	S2 How Set +1.20
G1CE	5	3	+5.00	S7 Group-spec +1.13
ME5X	3	5	+3.40	S3 Past Calls +1.00
BX	3	3	+3.33	S1 Cheat Sheet +1.00
CX50	2	4	+2.00	S1 Cheat Sheet +1.00
PDII	2	4	+1.75	S1 Cheat Sheet +0.75
GR50	5	11	+1.62	S1 Cheat Sheet +0.71
RE50	3	3	+1.33	S7 Group-spec −1.00
EN50	6	4	+0.92	S5 Personal +1.00
EL50	4	3	+0.67	S6 News +0.50
PR50	4	1	+0.50	S4 Relationships −1.00
ST50	3	4	+0.42	S1 Cheat Sheet +0.50
SC50	5	4	+0.10	S3 Past Calls +1.40
STII	4	4	−0.50	S3 Past Calls +1.50
FE5X	4	4	−0.50	S2 How Set −0.75
DF50	2	3	−1.50	S3 Past Calls −0.67
HR5X	6	3	−2.17	S8 Alumni −1.00
LE5X	0	6	ALL-LOST	Total = 10.3/16 (briefings score well; deals still lost)
TEII	0	5	ALL-LOST	Total = 7.0/16 (briefings ALSO score lower)

How to read this Positive Total Δ = won briefings score higher than lost briefings in this group → briefing craft is helping. Negative Total Δ = losing briefings actually score higher → briefing craft is overwhelmed by structural factors. All-lost groups = no matched wins in this sample (likely selection bias or community-level issue worth investigating).

§2 — Detailed cut: groups with W ≥ 4 and L ≥ 4

Where the sample sizes give us actually-comparable apples-to-apples reads.

Group	W / L	Total Δ	Top 3 differentiators
EN50	6 / 4	+0.92	S5 Personal +1.00 S2 How Set −0.67 S4 Relationships +0.33
FE5X	4 / 4	−0.50	S2 How Set −0.75 S7 Group-spec −0.75 S5 Personal +0.50
GR50	5 / 11	+1.62	S1 Cheat Sheet +0.71 S5 Personal +0.64 S4 Relationships −0.60
SC50	5 / 4	+0.10	S3 Past Calls +1.40 S7 Group-spec −1.05 S5 Personal +0.55
STII	4 / 4	−0.50	S3 Past Calls +1.50 S6 News −1.00 S2 How Set −0.50

Where briefing craft pays off

GR50 (Growth 50) — biggest loss volume but best craft-payoff: cheat sheet + personal intel discriminate strongly
EN50 (Enterprise 50) — personal intel is the lever (+1.00)
SC50 (Supply Chain 50) — annotated past-call history (+1.40) does the work; group-specific tailoring is over-applied

Where briefing craft is overwhelmed

HR5X (CHRO Division) — losing briefings score 2.17 points higher; alumni section bigger in losses (−1.00). MDAs over-research saturated communities.
FE5X (Finance Exec Division) — losing briefings have richer S2 (How Set) and S7 (Group-spec)
STII (Strategy 50 II) — same pattern: losses score higher overall

§3 — Thick-company effect

The single biggest predictor of close isn't briefing quality.

For each prospect, count the number of existing W50 members at their company. Bucket by density. Compute win rate within each bucket.

Existing members	Won n	Lost n	Win rate	Won total /16	Lost total /16
0 members	22	17	56%	8.73	6.24
1–2	11	22	33%	9.55	7.86
3–4	13	30	30%	8.46	9.03
5–9	14	22	39%	11.07	9.95
10+	25	8	76%	11.24	8.75

0 members

56%

n=39

1–2

33%

n=33

3–4

30%

n=43

5–9

39%

n=36

10+

76%

n=33

The Goldilocks-bad zone Companies with 3–4 existing members close at 30% — the lowest win rate of any bucket. Enough W50 presence to surface in research, not enough density to carry the deal. Either commit to the account (build to 10+ members) or accept that 3–4 is structurally the hardest position.

The 0-member surprise Pure new-logo prospects (0 W50 members at company) close at 56% — second-highest rate. These are the deals that ride entirely on personal narrative: founder-CEO loneliness, succession moments, first-time CEO. The briefing has nothing to lean on but personal intel, and when it leans hard there, it wins.

The 10+ thick-company effect Companies with 10+ existing W50 members close at 76%. The community is "inside" the org — the deal is mostly already won; the briefing's job is not to break it. Notably, lost briefings in this bucket have the lowest total scores of any bucket (8.75) — when these deals lose, briefing craft was poor. Wins, by contrast, score 11.24/16.

§4 — Reading the implications

What this changes about how we operate.

Investigate LE5X and TEII separately. Both communities show 0 wins in this matched sample. LE5X briefings score well (10.3/16) but still lose — strongly suggests structural factors (community fit, peer-org overlap, budget cycles) outweighing briefing craft. TEII briefings score lower (7.0/16) — could be either a quality issue or selection effect. Pull SF win-rate data unfiltered for both before drawing community-level conclusions.
HR5X and FE5X are saturated communities where briefing craft can't dig you out. When losing briefings consistently score higher than winning ones (HR5X −2.17 total), the implication is that MDAs are doing extra research on the wrong cases. Consider deprioritizing high-effort briefings on these accounts unless a clear trigger event is present.
For GR50 (Growth 50), the cheat-sheet quick-take + personal intel is doing real work. 16 matched briefings (5 won + 11 lost) — the largest loss-volume group in the sample. Of all communities, GR50 is where the operator-level briefing-craft fix has the largest absolute impact.
Prospects at 3–4 member accounts need extra craft. The Goldilocks-bad zone is where briefing quality matters most because structural advantages don't carry the deal. Lean into Habits 2 (cheat-sheet quick-take), 3 (annotated past calls), and 4 (specific personal intel) hardest in this band.
Don't rest on 10+ accounts. Even though win rate is 76%, the 8 losses in the 10+ bucket all had below-average briefing scores (8.75/16 vs winners' 11.24). Briefing quality is what separates the 24% lost from the 76% won when the structural advantage is large.