---
name: opportunity-evidence-control
description: "Load when a sales leader or RevOps user pastes a dated CRM opportunity export and wants each open deal judged on-track, stalled or at-risk with a specific next move, plus a pipeline anomaly brief covering coverage by owner, sandbagging candidates, forecast and record-integrity flags, and repeated failure mechanisms. Needs an as-of date and asks for one rather than guessing. Not for forecasting the number, rep scorecards, or win/loss retrospectives."
---

# Opportunity evidence control

The export arrives as pasted text. If it carries no as-of date, ask for it in one line before any time-based judgment.

## The operating model behind the review

An opportunity is a claim about a future commercial event. The CRM contains observations that
may support or weaken that claim, but no field proves it by itself. Read the book as an evidence
system: identify the claim, list the unresolved buyer decisions, rank the available evidence,
and choose the cheapest intervention that can falsify or strengthen the forecast this week.
This approach prevents both optimism bias and reflexive pessimism.

Separate evidence into four levels. Level 0 is seller narration: notes, hopes, internal plans,
and forecast labels. Level 1 is buyer attention: a reply, attendance, or acknowledged material.
Level 2 is buyer commitment: a named decision process, allocated people, agreed success
criteria, or a scheduled decision event. Level 3 is irreversible progress: approved budget,
completed security review, accepted commercial terms, or executable paper. Confidence should
rise with evidence level, not with the number of CRM updates. Several Level 0 records do not add
up to one Level 2 commitment.

Every opportunity should carry a falsifiable deal thesis with five parts: the buyer problem, the
consequence of leaving it unsolved, the person who can authorize change, the process by which
that person decides, and the next observable proof. When one part is absent, diagnose the gap
rather than filling it with a conventional sales assumption. A good coaching action tests the
weakest part of this thesis. If the economic consequence is unknown, quantify it with the buyer;
if authority is unclear, test access; if decision mechanics are unclear, build a mutual action
plan; if proof is missing, schedule the event that produces it.

Stage governance is evidence governance. A stage is credible only when its exit evidence exists,
not when a rep has changed a dropdown. Qualification should end with an agreed problem and a
viable buying path. Validation should end with accepted success criteria and a recorded outcome.
Commercial work should end with a buyer-reviewed offer and a known approval path. Contracting
should end with resolved exceptions, identified signers, and an executable sequence. Where the
label and evidence disagree, report the evidence state and treat the label as a data-quality
signal.

Forecast categories answer a different question from stage. Stage describes unfinished work;
forecast posture expresses confidence that the work will finish in the period. A late-stage deal
can be a weak forecast when buyer proof is absent, and an early-stage deal can have strong intent
without being closeable in the period. Never use one label to validate the other. Reconcile them
against customer evidence, remaining gates, and calendar capacity.

Time has three distinct meanings. Age measures how long a state has persisted. Silence measures
how long evidence has gone unrefreshed. Runway measures how much calendar remains for unfinished
work. These clocks must not be collapsed into one “stale” judgment. Old stage age with fresh
buyer proof may indicate a governance or process bottleneck. Short stage age with no buyer proof
may indicate premature advancement. Ample runway changes urgency but does not convert weak
evidence into progress.

Use base rates without pretending they are facts about one buyer. Larger transactions, multiple
approval functions, security work, procurement, and legal exceptions usually require more
calendar than a simple renewal or standard order. State this as feasibility reasoning and point
to the actual unfinished work. Do not invent a universal sales-cycle duration. The purpose is to
test whether the documented plan can fit, not to enforce an unsupported industry average.

Treat the next event as an experiment. It should have an owner, counterparty, date, intended
decision, expected artifact, and a branch for failure. “Schedule a call” is incomplete unless the
call has a decision purpose. “Send information” is incomplete unless receipt or review will be
verified. The strongest intervention converts a vague state into new evidence and precommits the
forecast consequence if that evidence does not arrive.

At book level, distinguish four management systems. Deal control asks whether an individual
opportunity can advance. Forecast control asks whether confidence matches evidence. Capacity
control asks whether enough plausible value exists to meet the target. Data control asks whether
the system of record can support any of those decisions. Name the system that is failing; a
coaching program cannot repair insufficient opportunity creation, and more pipeline cannot make
an impossible close date true.

Look for repeated mechanisms, not coincidental labels. A pattern is meaningful when multiple
rows fail through the same controllable mechanism: no access to authority, undefined validation,
buyer events without dates, late legal engagement, or repeated record drift. Report the count,
value, common mechanism, and management response. With a very small sample, call it an operating
signal rather than a statistical conclusion.

Prioritization combines value, forecast exposure, evidence weakness, reversibility, and time to
intervene. A large weak claim near a reporting boundary may need executive resolution; a smaller
record defect may need immediate hygiene but no executive escalation. Do not let ARR alone order
the work, and do not let a red status erase the difference between a recoverable risk and a claim
that must be removed from forecast.

The final artifact must be auditable by a reader who did not perform the analysis. Each material
conclusion should trace to a supplied field or a clearly marked absence. Calculations should show
their components. Recommendations should state who acts, what evidence is sought, by when, and
what changes if the evidence fails. Concision means removing repetition, not removing the facts
needed to verify the decision.

Treat the export as a measurement taken at one observation date, not as a story that must be
made coherent. The job is to decide what leadership can rely on, what is merely recorded, and
what must happen next. Use only supplied facts. An empty field stays unknown. A confident CRM
label never outweighs contradictory evidence.

## Establish the measurement contract

Find the reference date before doing elapsed-time work. If none is supplied, do not infer the
current date. Explain that age, silence, runway, urgency, and sandbagging cannot be measured
faithfully; request the date; then provide only date-independent help such as a proposed stage
crosswalk and static record defects. Do not issue statuses or time-dependent ratios on a guessed
clock.

At the top of a normal review, print a compact policy table. These are the defaults:

| Canonical stage | Maximum days in stage | Maximum days without activity |
|---|---:|---:|
| Discovery | 30 | 21 |
| Evaluation | 45 | 21 |
| Proposal | 21 | 14 |
| Negotiation | 14 | 14 |

The default sandbagging signal is the intersection of three facts: close is within 30 days,
silence is at least 14 days, and forecast is not Commit. A coverage shortfall is raw open ARR
below three times remaining quota. Mark every supplied amendment as an override and every
unmentioned value as a default. An amendment changes only its named cell or rule; it does not
spread to other stages or tests.

Map unfamiliar stage labels by unfinished buyer work. Qualification belongs to Discovery,
technical proof belongs to Evaluation, a priced commercial offer belongs to Proposal, and
contracting or redlines belong to Negotiation. Present the proposed crosswalk and ask for
confirmation rather than silently asserting that local terminology is standard.

## Read evidence, not labels

For every open row calculate days since activity, days in stage when entry date exists, and days
to close. Test four independent questions:

1. Is stage age beyond its policy limit?
2. Is verified engagement too old?
3. Is the next buyer event controlled, dated, and attributable?
4. Can the remaining work plausibly fit before close?

The fourth question can dominate the first three. Recent contact does not compress a buying
process. Discovery still needs qualification, proof, commercial agreement, and paper; Proposal
still needs negotiation; Negotiation still needs approvals and execution. A close date that
cannot contain the unfinished work makes the row at-risk even when the numeric limits pass.
State both facts so the reader sees why this is judgment rather than arithmetic drift.

A value on a boundary is evidence of no safety margin, not evidence of health. Say “at the
limit” when equality occurs. Combine it with runway and the buyer-controlled gate. A deal one
day inside a silence limit can still be at-risk when signature is only days away.

Activity quality matters. A seller sending a note is not the same as a buyer replying, attending,
approving, or signing. When the export cannot distinguish one-way activity from two-way
engagement, name that uncertainty. An old outbound message with no recorded response is weak
evidence and should not reset confidence merely because it occupies the activity field.

Evaluate the next event as a control instrument. It is strong only when the record identifies an
actor, an observable act or decision, and a date. “Awaiting,” “pending,” “under review,” “in
progress,” “TBD,” and a past appointment describe state but do not establish a future event.
An external committee, board, counsel, procurement queue, or executive approval has its own
calendar; obtain that calendar before trusting a close date. Commit resting on an undated buyer
action is at-risk regardless of recent seller activity.

Use these dispositions consistently:

- **at-risk**: close feasibility, integrity, missing value, buyer control, or multiple strong
  signals make the current forecast unsafe and require intervention now;
- **stalled**: progress or documentation is weak, but runway remains credible and no immediate
  forecast contradiction exists;
- **on-track**: timing, evidence, next event, and stage path are mutually consistent, with any
  watch condition stated explicitly.

Do not soften a row to improve the appearance of the book. Conversely, do not call every weak
phrase a crisis: severity depends on stage, value, forecast category, and remaining runway.

## Protect the population and arithmetic

Closed Won and Closed Lost rows are historical context. Exclude them from the forward coaching
population and open-pipeline numerator, while using them only for clearly labeled historical
patterns. A past close date on an open row is not ordinary aging: remove it from active coverage
until status is resolved, flag record integrity before coaching, and require a new evidence-based
date. Repeatedly pulling a date earlier while stage does not advance is an acceleration anomaly,
not the opposite of risk.

Missing data disables only dependent operations. Without stage-entry dates, disclose that dwell
cannot be tested for any affected row while still using activity, close feasibility, and buyer
evidence. A missing ARR prints as a dash, makes the deal at-risk for hygiene, excludes it from
value sorting and coverage, and never removes the deal from review. Without quota, report open
ARR qualitatively, request quota, and do not manufacture a ratio. Without probability weights,
label raw coverage as an optimistic ceiling rather than pretending it is weighted coverage.

Compute per owner before computing a team total. Show numerator, quota, and ratio so every sum
can be audited. Re-add the component deals before publishing. Keep stage-quality risk separate
from a coverage shortage: the former calls for deal intervention, while the latter often calls
for pipeline creation. Concentration in one fragile opportunity should be called out even if the
raw ratio appears adequate.

## Prescribe evidence-producing work

Coaching table columns: Deal, Rep, Stage, ARR, Close date, Days since activity, Status, Diagnosis, Next action. At-risk first by ARR, then stalled, then on-track summarised per rep.

A useful action changes what will be known this week. Name the rep, buyer role, action, deadline,
expected evidence, and escalation or forecast consequence. “Follow up” is not an action. Prefer
requests such as confirming a redline-return date, booking a decision meeting, obtaining a board
agenda date, or resetting close and forecast when proof does not arrive. Do not invent contact
names that are absent from the export.

Lead with the controlling issue, then give a complete compact table. Put at-risk rows first by
ARR, stalled rows next by ARR, and summarize on-track rows without hiding them. Include diagnosis
and next move. Follow with a brief that separates coverage, sandbagging, forecast mismatch,
record integrity, and rep or segment patterns. Use different shapes when the work differs: a
missing-date response is a blocker note plus crosswalk; a cohort failure deserves a cohort-led
brief; an integrity exception belongs before the coaching table. This keeps paired reviews from
becoming mechanical copies.

Before returning, reconcile row count, open ARR, owner sums, ratio math, boundary language,
override scope, closed-row exclusions, blank fields, status consistency, and whether every urgent
row has a dated intervention. Prefer a complete answer over decorative prose.