Measure engagement value

Airlift measures migration and modernization outcomes; it does not assume a fixed time reduction. The unit of evidence is a benchmark observation for one engagement, estate, activity, method, and optional object.

For every observation record four numbers separately:

• manualBaselineMinutes: hands-on time for the comparable manual process;
• assistedHandsOnMinutes: developer or operator time with Airlift;
• automatedRuntimeMinutes: unattended machine runtime;
• reworkMinutes: additional hands-on correction effort.

This prevents a long automated job from being presented as developer effort and prevents rework from disappearing inside a headline percentage.

Record observations

{
  "engagementId": "eng_...",
  "estateId": "est_...",
  "objectId": "obj_...",
  "activity": "modernization",
  "method": "time_study",
  "manualBaselineMinutes": 120,
  "assistedHandsOnMinutes": 45,
  "automatedRuntimeMinutes": 12,
  "reworkMinutes": 8,
  "confidence": "high",
  "dimensions": { "sourceSystem": "synapse", "objectType": "table", "complexity": "medium" },
  "evidenceRef": { "system": "airlift", "type": "time_study", "id": "study-42", "digest": "<sha256>" },
  "evidenceDigest": "<sha256>",
  "observedAt": "2025-06-15T12:00:00Z"
}

fa value record --file benchmark.json --idempotency-key "benchmark-study-42"
fa value observations --engagement-id "$ENGAGEMENT_ID" --json

Supported methods are historical_actual, time_study, expert_estimate, and paired_execution. A system-recorded paired execution must reference Experiments.

Mint a summary

Select the exact observation IDs included in the calculation:

{
  "engagementId": "eng_...",
  "benchmarkObservationIds": ["bmk_...", "bmk_...", "bmk_..."],
  "reason": "Summarize the accepted engagement time studies."
}

fa value summarize --file summary.json --idempotency-key "value-summary-current-scope"
fa value summaries --engagement-id "$ENGAGEMENT_ID" --json

Airlift computes hands-on minutes saved, hands-on reduction rate, acceleration multiple, sample count, methods, source systems, and confidence. A new benchmark observation makes every active summary for that engagement stale. Mint a new summary instead of silently changing a previously published result.

Publish safely

Client reports require at least three observations and medium or high computed confidence. Lower-confidence summaries can be published only for an account-team audience and retain their limitations.

fa value publish --file report.json --idempotency-key "value-report-client-1"
fa value reports --engagement-id "$ENGAGEMENT_ID" --json

Every generated claim identifies its engagement scope and evidence digest. Say “the recorded scope reduced hands-on effort by 58%,” never “Airlift always reduces migrations by 60%.”

Before using an aggregate claim outside one engagement, run fa delivery claim-check against a versioned cohort packet. The external gate requires at least three engagements, ten comparable observations, reviewed methodology, medium/high confidence, exclusions, limitations, and immutable evidence. See industrialize migration delivery.