Closeout report — Lead Assurance pilot (internal, controlled)
Prepared for: internal calibration · Status: illustrative sample — no executed run
1 · The frozen Outcome Contract
- Outcome verified
- For each dispatched test event, exactly one contact record carrying the exact, byte-for-byte event identifier appears in the designated CRM.
- Originating claim
- The producing system asserts "lead created" and emits a unique event identifier at dispatch time.
- Independent proof source
- The CRM itself, queried by a separate verifier identity holding read-only contact access. The producer and verifier never share credentials.
- Verification window
- 300 seconds, inclusive, from event dispatch.
- Settlement
- An additional 300-second settlement floor; a positive result requires at least 10 complete observations of the same exact singleton record through settlement, with mandatory re-audits at +5 and +30 minutes.
- Classification
- Deterministic and fail-closed: late or missing intended outcomes are failures; unavailable, contradictory, malformed, or duplicated proof remains uncertain; only events explicitly flagged as internal tests are excluded.
- Contract frozen
- Before the first counted event. No rule may change once execution begins.
2 · Planned run parameters (protocol specification — no run executed)
- Environment
- Specifies an isolated developer test account. No customer or production data would be used or accessible.
- Duration
- 10 operating days (~1.5 weeks). No day may begin until the prior day's pending events and audits reach zero.
- Event volume
- 250 controlled events, with failure modes to be injected deliberately so the classifier's honesty can be observed — not to simulate natural failure rates.
3 · Planned scenario mix and dispositions
| Planned scenario | Events | Disposition under the frozen rules | Why |
|---|---|---|---|
| Ordinary on-time outcome | 175 | Verified on time | Exactly one matching record inside the window, stable through settlement. |
| Late, missing, or wrong-ID | 35 | Verified failure | Intended outcome absent at deadline, or identifier did not match byte-for-byte. Lateness is not forgiven. |
| Proof source unavailable | 20 | Uncertain | The independent source cannot be observed through settlement. Nothing is inferred. |
| Duplicate records | 10 | Uncertain | The contract requires exactly one record; two would match. Ambiguity is preserved, never picked from. |
| Explicit internal test flag | 10 | Excluded | Only events explicitly flagged as internal tests are excluded — and each exclusion is itself recorded. |
These proportions are chosen, not observed in the wild. They demonstrate classification behavior only and say nothing about natural failure frequency.
4 · Illustrative event trace (how one on-time event would settle)
| T + | Channel | Observation |
|---|---|---|
| 0s | Claim | Producer dispatches the event; asserts "lead created"; identifier issued. |
| 212s | Evidence | Verifier (read-only) would observe exactly one matching record in the independent source. |
| 300s | Contract | Business deadline passes with the match present — inside the window. |
| 600s | Evidence | Settlement floor completes; ≥10 observations; singleton unchanged. |
| +5 min / +30 min | Audit | Mandatory positive re-audits recompute the result from preserved evidence — planned behavior, shown illustratively. |
Disposition the frozen rules would assign to evt_SAMPLE_0147
Verified on time
5 · Evidence the protocol preserves
- For every event, the protocol specifies sequenced, integrity-tagged evidence artifacts: dispatch record, each verifier observation, deadline evaluation, settlement observations, and final classification.
- Durable audit artifacts from the +5 and +30 minute recomputations, when a run is executed.
- Artifacts store payload digests and keyed integrity tags — never credentials or tokens.
- Artifacts are sequenced and tamper-evident. If later evidence contradicted a disposition, a linked superseding artifact would be created; an original entry is never rewritten.
Example artifact digest (synthetic): sha256:3b1f…9a04 — shown to illustrate format, not as verifiable data.
6 · Honest exclusions and what this sample does not establish
- An illustrative sample of the internal, controlled protocol. No run has been executed.
- Even a delivered report of this kind would not establish natural failure frequency, production security, customer value, willingness to pay, or generalization beyond its specific contract.
- In the planned allocation, the 20 proof-unavailable and 10 duplicate scenarios settle as uncertain and stay uncertain. Uncertainty is a deliverable, not an error.
7 · Evidence-integrity audit — illustrative
Independent recomputation of all preserved evidence — shown with an illustrative result on synthetic artifacts
Pass
This is the evidence-integrity audit result (PASS / FAIL / UNCERTAIN / UNAVAILABLE) — a separate layer from the outcome classifications in section 3. In a delivered report it would state that every preserved artifact independently recomputed and remained intact, and would prove nothing beyond that. Here it is illustrative: no artifacts exist to recompute.