deal_risk_review v1.0.0

name: deal_risk_review description: "Review one live B2B deal for stall and loss risk before it slips: score momentum, access, champion behaviour, process certainty, urgency, and no-decision pressure from evidence over time, triage it as save, downgrade, or exit, and run a dated save play for the pattern that is killing it. Use when a deal has gone quiet, pushed its close date, lost its champion, stalled in legal or procurement, or needs a hard call before a forecast review. For static MEDDPICC qualification scoring, use meddic_qualification instead."

Deal Risk Review

Purpose

Deals rarely die in one call. They decay: the next meeting never lands, the close date moves with no new information, the champion stops doing things, a paper step nobody started eats the last three weeks. This skill catches that decay while there is still time to act.

It answers three questions about one deal, in order:

1. Is this deal moving, stalling, or already dead?
2. What single pattern is doing the damage?
3. What do we do this week, by name and date, or do we downgrade or walk away?

Scope boundary: this is the intervention layer, not qualification. If the deal has never been qualified, run meddic_qualification first and bring its scores to this review. If the question is where a whole pipeline leaks across many deals, use sales_pipeline_leak_diagnosis. This skill takes one deal that was supposed to close and decides how to save it or stop spending on it.

Workflow

1. Assemble the evidence pack before anyone gives an opinion

Collect, from the CRM, email, and call notes. Do not score from memory.

• Stage, amount, first close date, current close date, and every date change in between
• Date and nature of the last buyer action (they did something, not you sent something)
• Every meeting in the last 30 days: who attended, by title, and what was decided
• The named economic buyer, and whether the seller has met them directly
• The named champion, and the last three things they actually did inside their company
• Decision steps and paper steps still open: security, legal, procurement, vendor onboarding, budget approval
• Any written next step with a buyer owner and a date. A seller task does not count
• Competitor, internal build, or do-nothing option, in the buyer's words if you have them

If half the pack is missing, that is the first finding. Missing evidence scores as risk in step 2, never as neutral.

2. Score six stall signals, 0 to 2 risk each

Score each signal from evidence only. Full rubric with evidence examples: references/scoring-rubric.md. Run it through the scorer so reviews stay comparable week to week:

bash
python3 ~/workspace/skills/deal-risk-review/bin/score_deal.py scores.json

| Signal | 0 Controlled | 1 Warning | 2 Critical | |---|---|---|---| | Momentum | Buyer action in the last 7 days, dated next step exists | Last buyer action 8 to 14 days ago, or next step has no buyer owner | No buyer action in 15+ days, or no dated next step at all | | Access | Seller has met the economic buyer and at least two other stakeholders | Economic buyer named, seller has not met them, champion claims access | Economic buyer unknown, or seller is single-threaded on one contact | | Champion behaviour | Champion did something costly in the last 14 days: booked a meeting, shared internal information, co-wrote the business case | Champion is responsive and positive but has done nothing costly yet | Champion went quiet, left, or failed a direct ask and gave no credible reason | | Process certainty | Decision and paper steps are listed, owned, and dated, working back from the buyer's go-live date | Steps are listed but undated, or paper process has not started | Decision or paper process is unknown, or a gatekeeper appeared late with new requirements | | Urgency | Buyer stated a dated consequence of not acting, in their words, tied to a real event | Pain is real but has no date, or the date is the seller's quarter end | Pain is vague, or the buyer said timing is flexible | | Competitive and no-decision pressure | Buyer named the alternatives and the criteria, and you know where you stand on each | A rival or internal build exists but standing is unclear | Buyer is avoiding a decision: options keep growing, meetings get cancelled, or the sponsor will not state a decision date |

The day thresholds are operating rules for this review, not research findings. A team with a 9 month enterprise cycle can widen them, but it must write the wider numbers down and use them consistently.

Total risk, out of 12:

• 0 to 3, Controlled: keep the plan, review again next week
• 4 to 6, Watch: one or two warnings. Fix them in the normal course of the deal, name an owner for each
• 7 to 9, High risk: a save play is required this week, and the forecast drops one category until a signal improves on evidence
• 10 to 12, Critical: assume stall or loss. Escalate to a manager review within 48 hours, run one save play with a hard deadline, and remove the deal from Commit

Compare with the last review if one exists. A deal that moved from 4 to 7 in two weeks is in more danger than a deal that has sat at 7 while a dated security review finishes. Direction matters as much as the number.

3. Name the killing pattern

Pick one primary pattern. A deal can have three problems, but a save plan with three fronts usually has none.

• Going dark: momentum 2, everything else was fine. Contact stopped responding.
• Coach, not champion: access or champion behaviour at 2. Your contact gives information and encouragement but has never spent political capital.
• No economic buyer: access 2. Nobody who can release the money has heard the case from you.
• No decision: urgency or competitive pressure at 2. The buyer likes the product and will not decide. Dixon and McKenna report in The JOLT Effect that 40 to 60 percent of deals end this way, lost to no decision rather than to a rival, so treat doing nothing as a named competitor in every review.
• Paper ambush: process certainty 2. Legal, security, or procurement started late, or a new gatekeeper reset the clock.
• Champion loss: champion behaviour jumped to 2 because the person left, changed roles, or lost standing.
• Fantasy timeline: urgency 2 and process certainty 1 or worse. The close date comes from the seller's quota, not the buyer's event.

4. Run the matched save play, with a deadline

Each pattern has one play in references/save-plays.md. Every play follows the same shape:

1. One diagnostic question to confirm the pattern before you spend effort
2. One buyer-facing move this week, small enough that a real supporter can say yes
3. One internal move: forecast change, manager escalation, or a second thread opened
4. A pass or fail test with a date. If the buyer does not do the thing by that date, the pattern is confirmed and the triage in step 5 applies

Worked example of a full review and save plan: references/worked-example.md.

Rules for save plays:

• Ask for behaviour, not reassurance. A meeting booked, a document shared, an introduction made. Compliments and positive replies are not advancement.
• Never run the same play twice on the same deal. If it failed once, the answer was information. Change pattern, escalate, or exit.
• Start paper steps in parallel with the business case, not after it. Security, legal, and procurement run on the buyer's calendar and do not speed up because your quarter is ending. Ask the buyer for their typical durations instead of guessing them.
• Multi-thread before you need to. If the only thread breaks, you have no deal, you have a pen pal.

5. Triage out loud: save, downgrade, or exit

Close every review with one verdict, recorded in the CRM the same day:

• Save: risk 9 or below, a play is running, pass or fail date is set
• Downgrade: risk 7 or above, or a play failed its test. Move the forecast category down and set a new close date from the buyer's process, working back from go-live, not forward from hope
• Exit: risk 10 or above with a failed play, or no economic buyer and no path to one after two direct attempts, or the buyer confirms there is no dated problem. Close it as lost to no decision or lost to competitor, honestly, so pipeline data stays useful. A clean exit frees the rep and keeps the door open. A zombie deal does neither

6. Run the weekly review in 30 minutes

Use the agenda and question bank in references/review-checklists.md. One deal per pass through steps 1 to 5. The manager's job in the room is to attack evidence, not effort: every claim in the pack gets the question "how do we know that, and when did the buyer do it?"

Output Contract

Every review produces one document with these sections, in this order:

1. Verdict: Save, Downgrade, or Exit, plus the risk total and its direction since last review
2. Evidence pack summary: last buyer action and date, close date history, who has been met by title
3. Signal scores: six scores, one line of evidence each. No score without its evidence line
4. Killing pattern: one pattern, named, with the diagnostic that confirmed it
5. Save play: buyer-facing move, internal move, owner for each, pass or fail test and its date
6. Forecast change: category and close date after this review, and the reason in one sentence

Keep it to one page. If a section cannot be filled, write Unknown and treat that gap as a scored risk, per step 1.

Operating Rules

1. Evidence beats confidence. A confident rep with no buyer actions in 15 days has a critical momentum score. The score follows the evidence, not the feeling in the room.
2. Unknown scores as risk. An unnamed economic buyer is not a neutral gap in a late stage deal. It is the deal's biggest exposure.
3. One pattern, one play, one deadline. Reviews that end with a list of five actions and no dates have decided nothing.
4. A failed test is data, not a reason to ask softer next time. Downgrade or exit on schedule.
5. Never invent buyer facts to complete the pack: no made up titles, dates, budget figures, or quotes. Write Unknown and score it.
6. Do not coach the rep to pressure a champion into promises they cannot keep, hide steps from procurement, or work around a gatekeeper. Save plays open doors. They do not cut lines, and a deal won by hiding the paper process usually dies in it anyway.
7. Record the verdict the same day. A review that lives only in a meeting is a forecast error waiting for quarter end.