Deal Qualification Gates: Evidence, Not Enthusiasm
The 2025 Ebsta x Pavilion benchmark put average B2B win rates at 19%. Most of that waste is not lost at the negotiation table; it is manufactured months earlier, when a conversation gets logged as a deal and nobody ever checks whether the evidence behind it hardened. CRM stages measure where a deal is. Almost nobody measures how WELL you know what you claim to know about it.
This skill installs that second axis: an evidence-quality score per qualification dimension, minimum scores per stage, and a standing qualify-or-kill discipline. Gates, not fields. Adding more CRM fields will not fix a forecast; minimum evidence per stage will.
The Evidence-Quality Scale
Score every qualification dimension 1 to 5. The scale grades evidence, never rep enthusiasm:
| Score |
Grade |
Meaning |
| 1 |
Unknown |
Never asked. |
| 2 |
Weak |
Surface mention, uncorroborated ("they said budget shouldn't be an issue"). |
| 3 |
Confirmed |
Buyer said it explicitly and it is logged (quote, transcript, email). |
| 4 |
Quantified |
There are numbers behind it (cost of the problem, headcount affected, deadline date). |
| 5 |
Validated |
Corroborated by a third party or system data (their CFO repeated it, usage data shows it, a signed evaluation plan names it). |
The jump that matters is 2 to 3: from "the rep believes it" to "the buyer said it and we can point at where." A pipeline where most dimensions sit at 2 is a mood, not a forecast.
The Dimensions
Default lens is SPICED, scored across six letters: Situation, Pain, Impact, Critical Event, Decision (criteria), Decision (process). Total out of 30. If your team runs MEDDICC or BANT, keep the scale and gates identical and swap the dimensions; the mapping table lives in references/qualification-frameworks.md. The framework is the vocabulary; the gates are the system.
The Stage Gates
Per-stage minimum TOTAL scores, calibrated for a six-dimension, 30-point model:
| Transition |
Minimum total |
What it enforces |
| Inbound to Discovery |
7 |
You know something real beyond a form-fill. |
| Discovery to Demo |
14 |
Pain and situation are confirmed, not assumed. Your demo has something specific to aim at. |
| Demo to Proposal |
19 |
Impact is at least quantified. A proposal without a number in it is a brochure. |
| Proposal to Commit |
23 |
Critical event and decision process are hard. You know why now and how they buy. |
Two enforcement rules:
- No single dimension below its floor. A deal can hit 19 total while Critical Event sits at 1; that deal does not move. Every dimension needs at least a 3 by proposal stage. One letter under its gate and the deal stays put. No rounding up. Go back, ask again, qualify or kill.
- The critical-event forcing question: "What breaks for this customer if they do nothing until next quarter?" No concrete answer scores Critical Event at 1, and the record is a conversation, not a deal. This single question deflates more zombie pipeline than any dashboard.
Why the paranoia about evidence depth: Gong's analysis of 1.8M opportunities found closed-won deals carry roughly twice as many engaged buyer-side contacts as closed-lost ones, and its call research found reps asking 11-14 targeted discovery questions correlate with the highest win rates (Gong Labs, 2017-2021 datasets). The same Ebsta x Pavilion dataset behind the 19% headline (655K opportunities, $48B pipeline, 2025 edition, still the latest full dataset as of August 2026) shows what the gates buy you: deals that close within ~50 days win at roughly 47% versus roughly 20% for deals that drag past that mark, and early economic-buyer involvement lifts win rates by around 55%. Both map directly onto the gates: qualify-or-kill is what keeps cycle time short, and the decision-process floor is what forces the economic buyer question early. Evidence quality is not bureaucracy; it is the observable difference between deals that close and deals that decay. Benchmark provenance and vintages: read references/qualification-frameworks.md.
Deployment
- CRM: one property per dimension (the 1-5 score) plus a computed total next to the account's ICP tier. Deal health = fit tier x evidence total, on one screen.
- Enforcement: stage-transition validation where your CRM supports it; otherwise a weekly below-gate exception report: every deal sitting above a stage its evidence does not support, with the missing dimensions named. The report is short or your pipeline is fiction, and either way you learn something.
- Qualify-or-kill review: each below-gate deal gets one of three verdicts, on a date: re-qualify (a named person asks the missing question by a named day), downgrade (back to the stage its evidence supports), or kill (closed-lost with the loss pattern recorded, see closed-lost-revival for what that record later earns you). "Leave it and hope" is not a verdict.
- Scoring hygiene: score at the moment evidence lands (after calls, from transcripts), not in a Friday batch from memory. If an AI agent processes your call transcripts, dimension scoring belongs in that pipeline; the transcript is the evidence, so the score should come from it, not from recollection.
- Expect two weeks of rep resistance, then adoption once the gates start protecting calendars from deals that were never going to close. The gate is not an audit of the rep; it is armor against wasted evenings.
Rollout Order
Week 1: score the CURRENT pipeline as-is, no consequences. The distribution is the diagnostic: a healthy pipeline shows scores rising with stage; a hope-based pipeline shows late-stage deals with early-stage evidence.
Week 2: install the gates forward-looking (new stage transitions only). Grandfather existing deals but flag them in the exception report.
Week 4: first qualify-or-kill review over the flagged backlog. Expect to kill 15-30% of "active" pipeline; the forecast gets smaller and true simultaneously, and that trade is the entire point.
What good looks like
- Evidence scores rise with stage. The week-one distribution is the diagnostic: late-stage deals carrying early-stage evidence is the finding, not an embarrassment to hide.
- Every score of 3 or higher points at something: a logged quote, a transcript, an email. The 2-to-3 jump is auditable, never vibes.
- No deal advances with any dimension under its floor, deals move backward when the evidence says so, and nobody treats a downgrade as a failure. A ratchet-only pipeline is the tell that the gates are theater.
- Every Commit-stage deal has a concrete answer to "what breaks for them if they do nothing until next quarter?" written in the record.
- The weekly below-gate exception report is short, and each deal on it carries a verdict with a name and a date: re-qualify, downgrade, or kill. "Leave it and hope" never appears.
- The first month kills 15-30% of "active" pipeline and the forecast gets smaller and truer at the same time. Leadership treats that as the win it is.
- Scoring happens when evidence lands, from calls and transcripts, not from Friday memory.
Diagnostic Questions
- Pick your three biggest deals. For each: what breaks for the buyer if they do nothing this quarter? If the answer starts with "I think," score it honestly.
- What percentage of proposal-stage deals have a quantified impact number IN THE BUYER'S OWN FIGURES anywhere in the record?
- When did a deal in this pipeline last move BACKWARD a stage? If the answer is never, stages are being used as a ratchet, and the forecast inherits the fiction.
- How many deals older than 2x your median cycle time sit in the pipeline, and what evidence score do they carry?
- Who is allowed to kill a deal, and when did that last happen without a manager forcing it?
1---2name: deal-qualification-gates3description: Install evidence-gated qualification on a live pipeline: score the QUALITY of evidence behind every deal (1-5 per qualification dimension), set minimum scores per stage, and enforce qualify-or-kill at each gate. Triggers on 'zombie deals,' 'pipeline is full of junk,' 'is this deal real,' 'qualify or kill,' 'deals stall at proposal,' 'forecast built on hope,' 'reps say every deal is closing,' 'stage criteria,' 'exit criteria,' 'MEDDIC scoring,' 'SPICED scoring,' 'qualification framework rollout,' or any situation where deals advance on rep optimism instead of buyer evidence. Works with SPICED by default and maps to MEDDICC or BANT. BOUNDARY: icp-builder owns ACCOUNT-level fit (is this the right company) and carries a summary of these gates as its pipeline-enforcement step; this skill is the full deal-level operating system (is this deal real). deal-velocity-engineer treats slow deals as a speed problem; this skill treats false deals as a truth problem, and it runs first, because accelerating an unqualified dea4---56# Deal Qualification Gates: Evidence, Not Enthusiasm78The 2025 Ebsta x Pavilion benchmark put average B2B win rates at 19%. Most of that waste is not lost at the negotiation table; it is manufactured months earlier, when a conversation gets logged as a deal and nobody ever checks whether the evidence behind it hardened. CRM stages measure where a deal is. Almost nobody measures how WELL you know what you claim to know about it.910This skill installs that second axis: an evidence-quality score per qualification dimension, minimum scores per stage, and a standing qualify-or-kill discipline. Gates, not fields. Adding more CRM fields will not fix a forecast; minimum evidence per stage will.1112---1314## The Evidence-Quality Scale1516Score every qualification dimension 1 to 5. The scale grades evidence, never rep enthusiasm:1718| Score | Grade | Meaning |19|---|---|---|20| 1 | Unknown | Never asked. |21| 2 | Weak | Surface mention, uncorroborated ("they said budget shouldn't be an issue"). |22| 3 | Confirmed | Buyer said it explicitly and it is logged (quote, transcript, email). |23| 4 | Quantified | There are numbers behind it (cost of the problem, headcount affected, deadline date). |24| 5 | Validated | Corroborated by a third party or system data (their CFO repeated it, usage data shows it, a signed evaluation plan names it). |2526The jump that matters is 2 to 3: from "the rep believes it" to "the buyer said it and we can point at where." A pipeline where most dimensions sit at 2 is a mood, not a forecast.2728## The Dimensions2930Default lens is SPICED, scored across six letters: Situation, Pain, Impact, Critical Event, Decision (criteria), Decision (process). Total out of 30. If your team runs MEDDICC or BANT, keep the scale and gates identical and swap the dimensions; the mapping table lives in `references/qualification-frameworks.md`. The framework is the vocabulary; the gates are the system.3132## The Stage Gates3334Per-stage minimum TOTAL scores, calibrated for a six-dimension, 30-point model:3536| Transition | Minimum total | What it enforces |37|---|---|---|38| Inbound to Discovery | 7 | You know something real beyond a form-fill. |39| Discovery to Demo | 14 | Pain and situation are confirmed, not assumed. Your demo has something specific to aim at. |40| Demo to Proposal | 19 | Impact is at least quantified. A proposal without a number in it is a brochure. |41| Proposal to Commit | 23 | Critical event and decision process are hard. You know why now and how they buy. |4243Two enforcement rules:44451. **No single dimension below its floor.** A deal can hit 19 total while Critical Event sits at 1; that deal does not move. Every dimension needs at least a 3 by proposal stage. One letter under its gate and the deal stays put. No rounding up. Go back, ask again, qualify or kill.462. **The critical-event forcing question:** "What breaks for this customer if they do nothing until next quarter?" No concrete answer scores Critical Event at 1, and the record is a conversation, not a deal. This single question deflates more zombie pipeline than any dashboard.4748Why the paranoia about evidence depth: Gong's analysis of 1.8M opportunities found closed-won deals carry roughly twice as many engaged buyer-side contacts as closed-lost ones, and its call research found reps asking 11-14 targeted discovery questions correlate with the highest win rates (Gong Labs, 2017-2021 datasets). The same Ebsta x Pavilion dataset behind the 19% headline (655K opportunities, $48B pipeline, 2025 edition, still the latest full dataset as of August 2026) shows what the gates buy you: deals that close within ~50 days win at roughly 47% versus roughly 20% for deals that drag past that mark, and early economic-buyer involvement lifts win rates by around 55%. Both map directly onto the gates: qualify-or-kill is what keeps cycle time short, and the decision-process floor is what forces the economic buyer question early. Evidence quality is not bureaucracy; it is the observable difference between deals that close and deals that decay. Benchmark provenance and vintages: read `references/qualification-frameworks.md`.4950---5152## Deployment53541. **CRM:** one property per dimension (the 1-5 score) plus a computed total next to the account's ICP tier. Deal health = fit tier x evidence total, on one screen.552. **Enforcement:** stage-transition validation where your CRM supports it; otherwise a weekly below-gate exception report: every deal sitting above a stage its evidence does not support, with the missing dimensions named. The report is short or your pipeline is fiction, and either way you learn something.563. **Qualify-or-kill review:** each below-gate deal gets one of three verdicts, on a date: re-qualify (a named person asks the missing question by a named day), downgrade (back to the stage its evidence supports), or kill (closed-lost with the loss pattern recorded, see closed-lost-revival for what that record later earns you). "Leave it and hope" is not a verdict.574. **Scoring hygiene:** score at the moment evidence lands (after calls, from transcripts), not in a Friday batch from memory. If an AI agent processes your call transcripts, dimension scoring belongs in that pipeline; the transcript is the evidence, so the score should come from it, not from recollection.585. **Expect two weeks of rep resistance,** then adoption once the gates start protecting calendars from deals that were never going to close. The gate is not an audit of the rep; it is armor against wasted evenings.5960## Rollout Order6162Week 1: score the CURRENT pipeline as-is, no consequences. The distribution is the diagnostic: a healthy pipeline shows scores rising with stage; a hope-based pipeline shows late-stage deals with early-stage evidence.63Week 2: install the gates forward-looking (new stage transitions only). Grandfather existing deals but flag them in the exception report.64Week 4: first qualify-or-kill review over the flagged backlog. Expect to kill 15-30% of "active" pipeline; the forecast gets smaller and true simultaneously, and that trade is the entire point.6566## What good looks like6768- Evidence scores rise with stage. The week-one distribution is the diagnostic: late-stage deals carrying early-stage evidence is the finding, not an embarrassment to hide.69- Every score of 3 or higher points at something: a logged quote, a transcript, an email. The 2-to-3 jump is auditable, never vibes.70- No deal advances with any dimension under its floor, deals move backward when the evidence says so, and nobody treats a downgrade as a failure. A ratchet-only pipeline is the tell that the gates are theater.71- Every Commit-stage deal has a concrete answer to "what breaks for them if they do nothing until next quarter?" written in the record.72- The weekly below-gate exception report is short, and each deal on it carries a verdict with a name and a date: re-qualify, downgrade, or kill. "Leave it and hope" never appears.73- The first month kills 15-30% of "active" pipeline and the forecast gets smaller and truer at the same time. Leadership treats that as the win it is.74- Scoring happens when evidence lands, from calls and transcripts, not from Friday memory.7576## Diagnostic Questions77781. Pick your three biggest deals. For each: what breaks for the buyer if they do nothing this quarter? If the answer starts with "I think," score it honestly.792. What percentage of proposal-stage deals have a quantified impact number IN THE BUYER'S OWN FIGURES anywhere in the record?803. When did a deal in this pipeline last move BACKWARD a stage? If the answer is never, stages are being used as a ratchet, and the forecast inherits the fiction.814. How many deals older than 2x your median cycle time sit in the pipeline, and what evidence score do they carry?825. Who is allowed to kill a deal, and when did that last happen without a manager forcing it?
swan-gtm