Lead-time audit — 2026-06-01

Summary

• Register size: 1163 actions
• Severity-4+ actions: 466
• Actions with first_press_mention populated: 46 (4.0% of register)
• Severity-4+ with first_press_mention populated: 46 (9.9% of sev-4+)
• Forward-cadence-eligible sev-4+ (filed within 14d of announced): 12 — structural ceiling on the gate-defensible cohort at current filing-cadence.

Coverage shortfall (Gate-3 publishability blocker)

The forward-cadence sub-slice is the gate's publishability cohort (all/sev-4+ slices are dominated by historical backfill — see Methodology). Until the forward-cadence sev-4+ slice reaches n=30, the headline percentage is not publishable externally. This section makes the gap a first-class statistic so the audit cannot silently read thin.

Bottleneck: forward-cadence filing rate. Only 4 eligible-but-unpopulated entries exist; even if wake-leadtime populated all of them, the forward-cadence sev-4+ cohort would reach only 12, still 18 short of the n=30 floor. wake-filing forward-cadence sev-4+ throughput is the binding constraint.

Operational read: wake-leadtime populates first_press_mention on existing actions, so it can only reduce the eligible-but-unpopulated count. wake-filing growing the eligible cohort itself is the only way to lift the ceiling once the eligible-but-unpopulated count is exhausted. Both wakes feed this number; neither alone is sufficient if the constraint is on the other side.

Headline metrics — all populated actions

• Sample size: 46
• Median lead time: +210 days (negative = filed before press, good)
• Mean lead time: +292.8 days
• Range: +1 to +1253 days
• Actions where IPTM was ahead of press: 0 / 46 (0.0%)

Severity-4+ slice (the gate's stated population)

• Sample size: 46
• Median lead time: +210 days
• Mean lead time: +292.8 days
• IPTM ahead of press: 0 / 46 (0.0%)

Gate-3 evaluation (target: ≥30% sev-4+ ahead AND mean lag ≤ +14 days)

• ≥30% ahead: ❌ (0.0%)
• Mean lag ≤ +14 days: ❌ (+292.8 days)

Forward-cadence sub-slice (filed within 14 days of announced_date)

Excludes historical-backfill entries (filed long after the event), which the all/sev-4+ slices above are dominated by. This is the population that actually tests our current operating cadence against the news cycle.

• Sample size: 8 (of which sev-4+: 8)
• Median lead time: +5 days
• Mean lead time: +6.9 days
• Range: +1 to +13 days
• Forward-filed actions ahead of press: 0 / 8 (0.0%)

Forward-cadence sev-4+ gate test (target: ≥30% ahead AND mean lag ≤ +14 days)

• ≥30% ahead: ❌ (0.0%)
• Mean lag ≤ +14 days: ✅ (+6.9 days)

Methodology

For each action file in docs/iptm/actions/, the audit derives two dates:

1. IPTM filing date — the date of the first git commit that added the action file (git log --diff-filter=A --reverse). This is the canonical moment the event entered the register. 2. First mainstream-press date — the optional frontmatter field first_press_mention: { date, url }. The url must point to a primary mainstream-press outlet (FT, Reuters, Politico, Bloomberg, WSJ, Nikkei, AP, AFP, Bloomberg-syndicated regional desks).

Lead time = filing − press, in days. Negative means we filed before the press; positive means the press was first. Actions without a populated first_press_mention are excluded from the metric (and counted in the not-yet-measured tally).

Forward-cadence sub-slice = the subset of populated actions where filed − announced_date ≤ 14d. Historical backfill (filed long after the event) is mechanically excluded. This is the population on which the gate's claim about operating-cadence vs. news-cycle is actually defensible.

Distribution (lead-time histogram, days)

Per-action lead times

Columns: Announced = action announced_date. Filed = first commit adding the action file. Press = first mainstream-press date. Lead = filed − press (negative = ahead of press). Fwd-d = filed − announced (≤14 marks the action as forward-cadence vs. historical backfill).

Honest caveats

• Backfill bias. Actions filed long after the underlying event (historical backfill ticks) dominate the all/sev-4+ slices and drag the headline number far above the register's current operating cadence. The forward-cadence sub-slice above is the defensible operating-cadence number; the all/sev-4+ slices are retained for audit-trail completeness only.
• Press-mention selection. "First press mention" is curator-assigned, not derived from a corpus search. Curator bias is possible; the field's URL is the audit trail.
• Outlet definition. "Mainstream press" means FT, Reuters, Politico, Bloomberg, WSJ, Nikkei, AP, AFP. Trade press (MLex, Inside Trade) is excluded by design — we are testing the register against the news cycle the buyer sees, not against specialist publications who already track this beat.
• Sample size. Until first_press_mention is populated on ≥30 severity-4+ actions, the headline percentage is not publishable. We report the number we have, and we report how much of the population is still unmeasured.