Deal risk scoring

The useful output is a risk flag on a committed opportunity, with the reasons named and cited. Silence, a stalled next step, or a missing buyer are reasons. A new close date, a pulled forecast category, and a stage change are not. The manager still owns the forecast.

Score this quarter's commit and best-case, not the whole open pipeline. Files that were never qualified belong with a disqualification recommender, not here.

Revenue intelligence and conversation tools in the Aviso, Clari, Gong, and Salesforce class all sit on this problem. This page does not rank them. Aviso and Clari Copilot are one class of inspection surface, not a bake-off. Gong holds the calls and emails you will quote. Salesforce is the opportunity of record. None of them should slip the deal for you.

Do not publish a forecast-accuracy lift. Yours depends on how honestly you commit, how long the cycle really is, and whether reps start managing the score.

Last-mile deal coaching names one next human action on a flagged deal. This page stops at the flag. Do not merge them into a workflow that both flags the opportunity and rewrites the commit.

Run the score only on opportunities the AE already put in commit or best-case. Refuse a flag that cannot cite decay, a MEDDIC gap, or both.

Activity decay is buyer-side, compared with the pattern that got the deal this far.

• Email: last inbound from a buyer, not last outbound from the AE.
• Meetings: last completed meeting with a buyer on the invite, and whether the next one is dated.
• Calls: last Gong-class recording with a buyer speaking. An internal prep call does not reset the clock.
• Committee: who still replies. One champion answering while finance, legal, and the economic buyer go quiet is decay of coverage. That hole should already show on buying-committee mapping.

Missing MEDDIC on a commit deal is a risk input, not a disqualify. Use MEDDIC auto-extraction for what was actually said, then flag:

• No cited economic buyer, or none appearing this cycle
• No decision process with a next step the buyer owns
• Champion last quoted weeks ago, with no new internal sell
• Metrics or criteria that were only inferred

Cite the flag: CRM field, value, and last-modified date; last inbound, with date and who; a call quote with speaker and timestamp, or "no buyer-side recording since [date]"; MEDDIC status of confirmed with cite, empty, or stale. A score without those lines is a color.

Tune by motion. A same-week reply bar that fits transactional close will flag every enterprise security review. Segment commercial vs enterprise, new logo vs expansion, and whether legal or security is a known gate.

Show the AE the flag and cites at the same time as the manager, or first. A manager-only view turns scoring into an interrogation. Reps pad activity and the decay signal dies.

Enterprise negotiation goes quiet for reasons that are still a process: security in flight, legal redlines with a date, a monthly budget committee, a champion on leave. If a buyer-owned next step is on the file and the date has not passed, silence is the process.

The failure is a velocity model trained on mid-market cycles. It treats "no meeting this week" as slip. The AE books a courtesy check-in to reset the score. The buyer stops answering. You manufactured the stall.

Write exemptions the model can read: a dated next step the buyer owns; a known gate already on the opportunity; a named blackout with a return date. "I will ping them Friday" is not an exemption. When the dated step passes with no inbound and no new recording, flag a stalled next step with the original date. Do not treat a quiet champion as departed unless you have a bounce, an out-of-office that says they left, or a peer who said they are gone.

The following is a made-up negotiate-stage inspection, not a measured program and not reported results.

An AE has Northline Logistics, a mid-market 3PL, in commit. Close date is month-end. Forecast category is Commit. Next step in Salesforce: "waiting on legal." Last AE outbound was yesterday. Last buyer inbound was eighteen days ago, the champion forwarding the MSA to counsel. No economic buyer on a Gong recording this quarter. MEDDIC: Champion confirmed, Economic Buyer empty, Decision Process "legal then VP signs" with no date.

A velocity-only score marks it red: engagement decay, negative tone on "this is taking forever on our side," missing buyer. Overnight, a workflow pulls forecast category to Upside and pushes the close date.

That write-back is the failure. The manager finds out in forecast that commit moved itself. The champion was frustrated with their own counsel, not walking away. The empty Economic Buyer was a real gap. The silence sat on a legal packet that had actually been sent.

What the flag should have been:

• Stalled next step. "Waiting on legal" has no buyer-owned date. Cite: next step last modified eighteen days ago, no inbound from counsel.
• Missing economic buyer. Empty field; last call roster is champion plus ops.
• Not a sentiment close-lost. "Taking forever" is a process complaint. Leave tone off the verdict.
• Not auto-slipped. Stage and Commit stay until the manager reviews.

The manager keeps commit for the week, sets a dated ask for a redline return date, and puts the missing buyer on the next working session. If that date passes with no inbound, the flag returns with a new cite. If the champion then says the VP will not engage this quarter, that quote is evidence. The manager pulls the category. The model did not.

Call and email tone is a weak input. Accents, second languages, dry enterprise email, and sarcasm confuse sentiment models. A clipped "received, we will review" from counsel is how legal writes. Do not treat a negative sentiment score as a dead opportunity. If you show tone, show the quote, not a color. It is optional color on a decay or MEDDIC flag, not a closed-lost reason.

RevOps can build the queue and the thresholds. RevOps cannot pull commit. The AE's manager owns the decision.

The flag must contain decay and/or missing MEDDIC, each with a cite or "missing"; whether a buyer-owned dated exemption is in force; and a place for the AE to rebut. It must not change Salesforce stage, forecast category, or close date; write Closed Lost from silence or tone; clear the next step to "slipped"; or email the customer a breakup.

The manager reviews in the ritual you already run: deal review, forecast call, or a Clari-class inspection view. They confirm, override with a written process reason, or send the AE for the missing buyer or date. Until then, the opportunity stays where the AE left it.

Probabilistic forecast roll-up can consume deal-level risk as an input to a range. It still does not rewrite commit. If both layers auto-push numbers, two models are arguing with the forecast the VP has to defend.

Start on current-quarter commit above the value managers already inspect by hand. Backtest against deals that actually slipped, and check when the flag would have fired. A last-week flag on a slip you already knew about is a report. Do not turn that backtest into an accuracy percentage for a vendor or for this page.

Measure flags the manager agreed with, written overrides, uncitable flags sent back, and any opportunity whose stage or category moved without a human. If that last count is not zero, turn the write-back off.

Aviso, Clari, Gong, and Salesforce will only show the activity and fields you load. The quality bar is a flag a manager can open, read the cites, keep a quiet enterprise deal in commit, and pull a real stall without the CRM slipping itself.