Deal risk review

Deal Risk Review

Purpose

Take a single deal and produce an honest health verdict: which specific risks are present, how severe each is, what to do about each, and whether the deal is red, yellow, or green. The goal is to catch a slipping deal before the forecast call, not after it pushes.

Inputs

• The deal: account, stage, amount, close date
• Who's engaged (and who isn't) — champion, economic buyer, blockers
• Activity history: last touch, last meeting, momentum over the last 2–4 weeks
• What you're trying to advance and the next confirmed step (if any)

Method

Run every deal against a fixed risk taxonomy. Don't grade on vibes — for each risk, mark present/absent, score severity, and attach a mitigation. Then roll up to a verdict.

The seven risks (taxonomy):

#	Risk	Present when…
1	Single-threaded	Only one contact engaged; no second relationship in the account
2	No economic buyer engaged	You've never met or confirmed who controls the budget
3	No confirmed next step	No future meeting/milestone on the calendar with a date
4	Stalled / aging	No meaningful two-way activity in 14+ days, or deal age >> sales cycle for its stage
5	Active competitor	A named competitor is in the eval and not neutralized
6	Close-date slipping	The close date has moved out one or more times
7	No quantified pain	Pain is acknowledged but not tied to a number (cost, time, revenue, risk)

Score each present risk 1–3:

• 1 (minor) — present but contained; a clear, cheap action fixes it.
• 2 (material) — credibly threatens the close date or amount; needs a deliberate play.
• 3 (critical) — can kill the deal on its own; nothing else matters until it's addressed.

Mitigation per risk (one specific play each):

1. Single-threaded → name a second stakeholder to reach this week; ask your champion for a warm intro to one peer or their boss.
2. No economic buyer → propose an exec-aligned next step (ROI review, exec readout); use a "who else signs off?" question to surface them.
3. No confirmed next step → before any call ends, book the next one on the calendar; if none exists now, send a calendar hold with an agenda today.
4. Stalled / aging → send a pattern-interrupt (status of the business case, a deadline, or a permission-to-close-the-file email); if no response, downgrade forecast.
5. Active competitor → run a differentiation play on the buyer's stated criteria; arm the champion with a battlecard; trap-set on where the competitor is weak.
6. Close-date slipping → rebuild the mutual close plan backwards from a real compelling event; if there's no event, the date is fiction — re-date it honestly.
7. No quantified pain → run a cost-of-inaction calc with the buyer; convert pain to $ / hours / risk so the business case survives procurement.

Verdict roll-up:

• RED — any single risk scored 3, OR three+ risks scored 2. Deal is not commit-able; either intervene hard this week or pull it from the forecast.
• YELLOW — one or two risks at severity 2, no 3s. Closeable but exposed; mitigations must land before the date holds.
• GREEN — only severity-1 risks (or none). Healthy; keep momentum, don't over-manage.

Output shape:

VERDICT: <RED | YELLOW | GREEN>
Top risks:
- <risk> (severity N) → <one-line mitigation, owner + this-week action>
- <risk> (severity N) → <…>
Forecast call: <commit | best-case | omit> — <one-sentence why>

Tool binding

This skill runs from a verbal deal description alone. It gets sharper when connected to your stack — strongest with Doris, the reference integration, because the risk signals come straight from conversation evidence instead of your memory.

With Doris (recommended)

If the Doris MCP (mcp.meetdoris.com) is connected, resolve the deal and let the ontology supply the risk signals directly:

• ontology_resolve("deal", id, expand=["risks","objections","commitments","activity","close_date_changes"]) — risks gives the system's own risk read, activity reveals stalled/aging and last-touch, close_date_changes proves date-slip, commitments shows whether a confirmed next step exists, and objections surfaces competitor and pain gaps.
• search_transcripts(...) to quote the exact moment a competitor was named or a close date got hedged — evidence beats assertion in a forecast review. Prefer the ontology's evidence-backed signals over re-deriving risk from recollection.

With a CRM / CI / email MCP

• CRM MCP (Salesforce/HubSpot) → pull stage, amount, close-date history, contact roles, and last-activity timestamps for risks 1–6.
• Conversation-intelligence MCP (Gong/Chorus/Fireflies) → confirm who's actually engaged, whether a competitor is named, and whether pain was ever quantified on a call.
• Email MCP → measure last-touch and reply latency to score stalled/aging.

With nothing connected

Ask the user to describe the deal: account, stage, amount, close date, who's engaged, last touch, and the next confirmed step. Walk the seven-risk checklist out loud with them, score each risk, attach the matching mitigation, and output the verdict block. No integration required — the taxonomy is the engine.

Works without Doris

Fully functional from a spoken or written deal description — Doris only removes the recall step and replaces "I think" with transcript- and CRM-backed evidence for each risk.

Common mistakes

• Grading on optimism instead of running the full checklist — every deal gets all seven checks.
• Calling a deal green because the champion is excited while the economic buyer is a stranger.
• Listing risks with no mitigation, or a mitigation with no owner and no this-week action.
• Trusting a close date that has already slipped without re-anchoring it to a real compelling event.