Decision under review: whether predictive scoring is the smallest defensible next investment for a professional-services sales workflow.
Evidence boundary: This is a synthetic teaching example. It is not client work, a testimonial, or a claimed commercial result. Its purpose is to show the structure and evidentiary discipline of an M87 deliverable.
Proof-basis key
ReportedObservedCalculatedInferredUnknown
Evidence register
Basis
Material statement
Decision effect
Reported
Leadership believes inconsistent follow-up is causing qualified opportunities to be lost.
Establishes the concern, not its cause.
Observed
No organization-wide definition of a qualified lead is present in the supplied workflow.
Scoring labels cannot yet be evaluated consistently.
Observed
Lifecycle-stage and first-response timestamps are incomplete or applied inconsistently.
The baseline cannot currently support attribution.
Inferred
Automating a disputed definition would scale inconsistency rather than resolve it.
Predictive scoring is premature.
Unknown
The conversion effect of response time, lead source, and qualification criteria.
No defensible value claim is available yet.
Competing explanations
A · Insufficient lead volume
Not established. Volume by eligible source and period is incomplete.
B · Slow first response
Plausible, but timestamps are too incomplete to quantify the effect.
C · Qualification ambiguity
Supported. Teams do not share an enforceable definition or required evidence.
D · Missing predictive model
Not supported. Model readiness depends on stable labels and measurable outcomes.
M87 STUDIO
Representative Decision Packet · 02/02
Governing constraint
The workflow lacks a stable qualification and measurement contract.
The immediate constraint is not the absence of AI scoring. It is the absence of agreed labels, required evidence, and complete timestamps needed to determine whether any scoring mechanism improves the decision.
Decision options
Option
Decision
Reason
Build predictive scoring now
Do not authorize
Training and evaluation labels are not stable.
Buy a scoring product
Defer
A product would inherit the same measurement defect.
Standardize qualification and instrumentation
Recommend
Creates the minimum evidence needed for the later build decision.
Take no action
Retain as fallback
Valid if the decision owner will not enforce the measurement contract.
Smallest defensible next decision
Authorize a four-week measurement phase: one qualification definition, required evidence fields, first-response and stage timestamps, and a weekly contradiction review.
Measurement contract
Primary outcome
Qualified opportunities advancing to an agreed next stage.
Leading indicator
Eligible records meeting the evidence and timestamp contract.
Guardrail
No automated rejection, routing, or external communication.
Source of record
The governed CRM export approved for the review.
Value range
Unknown No savings, conversion lift, or revenue range should be claimed until the baseline is complete and the attribution method is approved.
Authorization boundary
Authorized by this decision
Define qualification evidence
Instrument required timestamps
Run the bounded measurement phase
Return a comparison packet
Still unauthorized
AI model procurement or build
Automated lead rejection or routing
Contact enrichment or outreach
Claims of commercial value
Next gate
After four weeks, compare completeness, stage progression, contradictions, and outcome coverage. Only then decide whether deterministic rules, AI assistance, or no build is warranted.