Pre-call briefings, the calls they prepare for, and the deals that close — what the data tells us when we read all three together.

§1 — What we looked at

Three data sources, paired by Opportunity, scored independently.

For each of 195 matched prospect Opportunities — drawn from a 12-month window of won and lost deals across multiple W50 communities — we paired three sources:

All scoring was done by an LLM-based agent reading source material directly, with explicit rubrics and a 0–2 ordinal scale per dimension. Briefing scoring and call scoring are independent (separate agent passes, separate rubrics) so the call score is not contaminated by knowledge of briefing quality.

How the workflow actually runs (so the analysis frames findings correctly): An MDA does outreach and books the call. The MDA + research coordinator prepare the briefing using public sources (LinkedIn, press releases, prior touchpoints). The SD takes the call. The MDA listens in and, after the call, records what was learned — including the prospect's stated objection — into the Salesforce Description timeline. For multi-call deals, the next briefing builds on that updated timeline. This means the post-call narrative reflects what surfaced live, not what should have been in the briefing.

§2 — The headline

Briefing quality has a hard floor effect. Above that floor, what matters is whether the SD uses the briefing on the call.

Two findings dominate everything else and reproduce on multiple slices of the data:

1. Bad briefings can't be saved. Briefings scoring 0–5 on the rubric (n=16) closed at 12% — versus 48% (n=54) for "decent" briefings (6–8) and 58% (n=155) for "good ones" (9+). The largest jump is at the floor: going from "bad" to "decent" quadruples win rate. Going from "decent" to "good" adds only 10 more points.
2. Above the floor, call execution is the dominant signal. On a 12-point call-utilization rubric, won calls average 6.78; lost calls average 4.49 (n=120 won, n=108 lost). That +2.29 spread is meaningfully larger than the won-lost spread on the briefing rubric itself (+1.27 on a 16-point scale). The briefing sets the ceiling; the call decides whether you reach it.

Together those tell a clean story: get the briefing above the floor, then make sure the SD actually uses it on the call. Everything that follows is a more specific cut on what "uses it" means in practice.

§3 — Five patterns we have high confidence in

Each pattern reproduces on a sample large enough to support the claim. Sample sizes are quoted inline.

§4 — A worked example

One transcript, three failure modes — and what each one tells us about where the briefing system actually breaks.

The patterns above are aggregate. To see how they show up in a single call, we walk through the Patrick O'Keefe / Universal Corporation deal. It's a representative case — not cherry-picked — and it lights up three distinct failure modes in one conversation.

This case is the operational map. The briefing system today addresses none of the three failure modes systematically. The next section proposes how to.

§5 — Three places where briefings can get sharper

Each maps to one of the three failure modes above. Each is a workflow change, not a tooling project.

5.1 — A pre-call research checklist for patterns we know surface live

From the 11 deals where the post-call narrative named a trigger that wasn't in the briefing, the dominant pattern is leadership churn. Five of the eleven are some flavor of "C-suite or near-C-suite transition in the last 12 months" — and four of those five are losses. This is the highest-leverage thing for the MDA pre-call research workflow to add, because it's checkable from public sources.

5.2 — Disclosure-status tagging on every flagged trigger

Patrick O'Keefe's case showed that knowing the trigger isn't enough — the SD also needs to know how confidently to use it. A trigger flagged as "publicly announced via 8-K" can be opened with directly. A trigger flagged as "inferred from LinkedIn change" needs a softer reference. A trigger flagged as "internal CRM note from prior call" generally shouldn't be referenced at all unless the prospect raises it first.

The proposal is to add a disclosure-status field to every trigger surfaced in the briefing:

This is a small Conga-template change — one new field per trigger entry, three controlled values — but it solves the "I wasn't sure what to under-share, over-share" failure mode directly.

5.3 — A live-pivot habit when the prospect surfaces novel context

Some triggers can only surface live — referrals the prospect names, internal politics they reveal, recent events that hadn't filtered into public sources at briefing time. The data shows the SD's response to live novel context matters as much as their use of pre-prepared triggers. In the 13 cases where a trigger surfaced only on the call (no briefing flag) and the SD substantively used it, the deal won 77% of the time (n=13 — directional only at this cell size, but consistent with the broader U3 pattern).

§6 — What we're not yet confident in

Findings the data hints at but the sample sizes don't yet support.

In the spirit of saying only what we can defend, here are the places where the directional pattern is interesting but the n is too small for us to recommend acting on it.

• Specific killer-matrix cells with n < 10. The Corporate × U3=2 cell (n=7), Both × U3=1 (n=8), and None × U3=1 (n=8) cells all show win rates that fit the broader pattern but shouldn't be quoted as specific percentages. The shape of the matrix is robust; the specific cell numbers are not.
• Whether the Corporate vs Relational trigger flavor distinction matters for outcome. Corporate-only wins 52% (n=23) and Relational-only wins 55% (n=109). At these sample sizes that gap is indistinguishable from sampling noise. We've kept the flavor distinction because it changes the SD's opening line (corporate triggers are referenced directly; relational triggers lead with the human), but we cannot claim one flavor wins more than the other.
• Whether the Description-revealed trigger pattern (10 of 11 reclassifications were lost deals) reflects MDA workflow gaps or the simple fact that lost deals surface novel context more often. Both interpretations are plausible. Distinguishing them requires data we don't yet have — namely, what the briefing standout said at the time it was written, not what's in Salesforce now. Until we instrument that, we cite the pattern descriptively, not causally.
• SD-level and MDA-level performance attribution. Some SDs and some MDAs likely produce above-mean utilization scores systematically; we have not run that cut. With ~10 SDs and ~10 MDAs in the cohort, individual-level cells get small fast and require careful handling. This is a high-priority follow-up.
• The specific lift from the proposed pre-call research checklist. We can show the patterns are researchable. We cannot yet show how much win-rate lift instituting the checklist would deliver. That requires forward-running a pilot and comparing checklist-using briefings to control briefings on similar deals.

§7 — Operational implications

Five concrete asks, ranked by leverage and ease of implementation.

§8 — One paragraph for the CEO

The summary if you read nothing else.