What the briefing produces.
What the call actually does with it.

§1 — Methodology

How we got from briefing rubric scores to call-level utilization to trigger-type classification.

Salesforce stores call transcripts (Otter/Gong-style auto-transcribed) in the Description field of each Event linked to a prospect Opportunity. We pulled all Events for our 184 matched cohort, filtered to Prospect Calls with populated transcripts, and joined them to the briefing rubric scores from v2.

The 6-dimension utilization rubric

Each transcript was scored 0–2 on whether the SD used what the briefing's parallel sections (S1–S8) had supplied.

Note on Description fill rate: 35% of Prospect-Call Events had no transcript saved. Otter integration is producing transcripts; they're just not making it to Salesforce reliably. Our analysis covers the 65% that did.

Known limitation — nickname false negatives in U2 (champion mention). The U2 score relies on the agent matching the briefing's named champion to references in the transcript. After a sales-team review surfaced this risk, we ran a systematic two-pass nickname-aware re-scorer across all 228 scored transcripts.

Pass A (high-confidence, n=41 briefings with extractable champion names): built a 70+ entry first-name → nickname dictionary (Donald→Don, Robert→Bob/Rob, William→Bill/Will, Elizabeth→Liz/Beth, Margaret→Maggie/Peggy, etc.), word-boundary-matched each variant in every transcript, and filtered contractions ("don't" ≠ "Don"). Of the 41 named-champion briefings, only 2 yielded confirmed false negatives — both already flagged by the sales team: Robert Dro / Donald Allan (SD said "Don" once in passing, re-scored 0→1) and James Barnes / Eric Woodward (SD said "Eric recommended that we speak," re-scored 0→2).

Pass B (exploratory, n=57 briefings where rubric stored "Yes-name" as a placeholder): proper-noun extraction from briefing Standout / Trigger / sf_reason_explanation fields produced 1 review candidate — Birgit Boykin, where the prospect spontaneously said "say hi to Joe and Elise" near end-of-call. On inspection the SD's own utterances do not include "Joe" or "Crowly," so the original U2=0 stands; the prospect volunteered the name without the SD using it as social proof.

Net impact of the full systematic pass: 2 corrections, both already applied. U2 won–lost delta: +0.32 → +0.27 (direction stable, magnitude –16%). Champion-mentioned vs not-mentioned win-rate gap: 8 points → 1 point. This strengthens the v2 finding that simply naming the champion isn't the lever; qualitative use as social proof tied to the trigger is. The systematic pass is the cleanest validation we have that the U2 score is largely robust to nickname noise — the manager's spot-check accounted for nearly all of the recoverable error.

§2 — Headline finding

Call utilization is a bigger predictor of outcome than briefing quality itself.

§3 — The six utilization dimensions, ranked by delta

Trigger events are the single largest miss.

Champion utilization (U2) is high in both won and lost calls (1.69 vs 1.42) — the SD is mentioning the champion in most cases. Personal hook (U1) is also relatively well-utilized (1.70 vs 1.21). Where the gap explodes is on triggers — won calls cite trigger events at 3.3× the rate of lost calls. (Note: U2 was re-scored after a sales-team review surfaced nickname false negatives — e.g., the SD said "Don" but the briefing named "Donald Allan." Two confirmed cases were corrected; +0.32 → +0.27. A subsequent systematic two-pass nickname re-scorer over all 228 scored transcripts found no further false negatives — see §1 methodology for details.)

§4 — The trigger-event gap

Briefings flag triggers in 85% of cases. SDs surface them on calls only on the won side.

The MDA does the work. The briefing surfaces the trigger. The SD doesn't reference it on the call in 84% of lost cases. Single biggest fixable gap in the data.

§5 — Trigger taxonomy: two flavors

"Trigger" is shorthand for the credible "why now?" hook. It comes in two distinct flavors.

The Freedman / Jefferies case (twin brother is a GC50 alum) doesn't fit any of the corporate triggers (M&A, new role, etc.) — but it functions as a trigger on the call because it answers "why now?" The taxonomy splits cleanly into two flavors that work the same way operationally:

Distribution and outcomes (n=228)

Two patterns emerge:

• Most prospects (54%) have a Relational trigger only — referrals, mentors, family connections, OK-to-mention peers. Pure Corporate triggers without any relational hook are rare (7%).
• "Both flavors stacked" wins highest (67%) — when a corporate event (M&A, new role) AND a relational hook (named referral, mentor) are both present, the deal closes. This is the warmest profile in the data.
• "None detected" cohort closes at 29% — half the rate of prospects with any kind of trigger. Trigger presence matters more than trigger type.

§6 — The killer matrix

Trigger type × utilization. The single most actionable cell in the analysis.

Combining trigger classification with U3 utilization (whether the SD substantively surfaced the trigger on the call) produces the cleanest behavioral fingerprint of won deals in the dataset. Read this as a fingerprint, not a treatment effect — when an SD surfaces a trigger substantively, the prospect is usually warm enough to engage with it, so part of what these cells measure is prospect warmth bleeding through, not pure SD execution lift. The actionable part lives in the handoff gap, not the headline number — see callout below.

Three patterns leap out

1. Won calls cluster in the U3=2 row across every trigger type — Corporate 100% (n=6), Relational 83% (n=18), Both 83% (n=12). Treat that band as the shape of a won call, not as a forecasted lift; the smaller cells (Corporate n=6, Both U3=1 n=3) are noisy, and at least some of the height in this row is prospect warmth, not SD effort. The number to actually internalize is the row pattern, not the cell values.
2. Trigger exists but SD doesn't surface it (U3=0): 42–43% — this is the most useful row in the table. Across 81 calls (Corporate n=7, Relational n=60, Both n=14) the MDA flagged a trigger in the briefing and the SD walked away from it on the call. This is the leak the analysis is actually about — independent of any debate over the U3=2 row, MDAs and SDs disagree on which calls have a usable trigger ~32% of the time, and the win rate when they disagree is half what it is when they don't.
3. No trigger AND SD doesn't surface one (None + U3=0): 5% — 22 cases, one deal. When neither side of the handoff finds a hook, the call is essentially dead on arrival.

Examples

Two caveats — read this matrix as a fingerprint, not a forecast.

(1) Endogeneity / prospect-warmth confound. U3=2 ("substantively surfaced") requires the prospect to engage with the trigger reference for more than a sentence. Cold prospects shut these references down inside 30 seconds, which forces the SD to pivot away — and that call gets coded U3=1 or U3=0. So part of what the U3=2 row measures is the prospect's warmth bleeding through into the SD's score, not pure execution lift. The right read is "won calls look like this; lost calls don't" — a behavioral fingerprint — not "do this and your win rate jumps to 90%."

(2) Heuristic classifier + small cells. Trigger type is keyword-detected (briefing standout + sf_source + Champion field). It under-detects edge-case Relational triggers — the "None + U3=2 = 71%" bucket likely contains ~14 cases where a real trigger was on the call but my classifier missed it. The Corporate row (n=16 total) and the Both × U3=1 cell (n=3) are noisy enough that we should not literally quote the cell percentages; the row directions hold.

The robust finding that survives both caveats is the +0.74 U3 utilization delta across all 228 transcripts (Won 1.06 vs Lost 0.32, n's of 113 and 115) and the 32-point MDA→SD handoff gap (85% of briefings flag a trigger; only 53% of calls surface one). That's the part of this section that is not subject to small-cell noise or warmth confounding.

§7 — The soft-yes-fade is validated at the call level

The five phrase families I diagnosed in v1 from briefing prose now have actual call-level frequencies — and the multipliers are stark.

The four "decision-deferral" phrase families (details / circle / think / internal) all show 1.8–2.7× higher frequencies in lost calls. The "busy" family is much weaker as a signal — likely just generic exec stress, not a buying signal.

What this enables operationally: SDs can now be trained to recognize these phrase families in real time on the call and respond with deliberate counter-moves rather than letting them slide. Every "send me the details" is a 60% likelihood of loss if not addressed.

§8 — Briefing → utilization → win-rate ladder

The bottom of the rubric is the killer. Stamping out 0–5 briefings is the highest-leverage operational move.

Going from 0–5 to 6–8 briefings quadruples win rate (12% → 48%). Going from 9–11 to 12+ adds only 1 percentage point. The downside risk is at the bottom of the distribution, not the upside opportunity at the top. A "no Conga-only briefing leaves the door" policy would deliver more value than every subsequent briefing-craft optimization combined.

§9 — Trajectory at end-of-call is a near-perfect outcome predictor

The agent-coded trajectory label is essentially a real-time SD coaching signal.

"Positive-open" is the highest-volume recoverable category (76 calls, 26% win rate) — these are the deals where soft-yes-fade often appears. "Neutral" exits are the most expensive misclassification — only 4% recover, but reps frequently treat them as warm and continue nurturing for months.

§10 — The v2 endogeneity caveat is retired

Briefing craft is causal. Not because briefings produce magic, but because they produce raw material that effective SDs use on calls.

v2 honestly flagged that briefing quality and prospect quality are coupled. v3 breaks that confound. The chain is now observable:

1. Briefing rubric (0–5 → 12+) → Call utilization (2.94 → 8.09) — monotonic relationship
2. Call utilization (low → high) → Win rate (12% → 59%) — also monotonic
3. Trigger type × utilization matrix shows the cleanest causal pattern: same trigger, same prospect class, but different SD execution → 5% win rate vs 100% win rate

If briefings were just window-dressing for already-warm prospects, win rates would be similar across briefing buckets and utilization would be flat. They're not.

What we still can't conclude

• Whether the same MDA effort would change outcomes for prospects with no clear trigger (the "None" bucket would need a controlled experiment to settle).
• Whether SDs trained to surface triggers more aggressively would actually produce different utilization scores. Counterfactual is unknowable without an A/B test.
• Whether the 15 LOAD_FAILED summary-only transcripts reflect a sub-population of low-engagement calls — possible selection bias.

§11 — Operational implications

Six interventions: three for MDAs, three for SDs.

For the MDA team

For the SD team

Process recommendations

• Update the Conga briefing template with a structured Trigger Event field: dropdown (Corporate types / Relational types / Both / None) plus free-text specifics.
• Mandatory boolean on every Event: "Trigger event surfaced on call?" Yes / No. Pair with the briefing's Trigger field for direct audit.
• Salesforce dashboard: briefing-rubric → utilization → win-rate by SD. Coaching becomes data-driven. Trajectory categorization adds real-time pipeline health metrics.
• Pilot trajectory categorization on the next 30 days of calls — confirms the 95% closing rate and 4% neutral-rescue rate replicates outside this sample.

§12 — One paragraph for the CEO

What's still ahead. Future analyses worth running: (1) call-quality scoring rubric to mirror briefing rubric; (2) which specific MDAs and SDs consistently produce above-mean utilization (individual-level coaching); (3) lost-deal recovery sequences for "positive-open" trajectory deals (the 76-call recoverable cohort, 26% baseline win rate); (4) refining the trigger classifier with an LLM-based pass to clean up the "None + U3=2 = 71%" edge cases.