Sort Intelligence — operational constants, fidelity, and the three-role scan chain.

The fidelity cascade runs top to bottom. Each level's failures originate at the level above it:

Missorts and LIBs are Level 3 consequences. They originate at Level 1 — whether the sorter knows where the package goes before touching it. LTC (Label Training Certification) addresses Level 1 explicitly; GEMS #1760 addresses Level 0; the Sort Intelligence framework binds them together with empirical fidelity measurements at L2 and L3. *OJS / GEMS facts are internally sourced from facility records and not independently verifiable by an outside reviewer.

Four structural constants are derived from the corpus. Each is cross-validated across SOR/TMS and iGate/SEAS independently, and embedded as a live parameter in the tracker's OR models. They are governed by a single source of truth, metric_contract.json.

The κ_Z3 day-of-week vector on the SEAS basis is Mon 0.413 · Tue 0.395 · Wed 0.385 · Thu 0.380 · Fri 0.322 (full-year SEAS Hub Monthly mean ≈ 0.368 ± 0.014). These are small-sample estimates (≈ 5–6 sorts per day of week; Fri n = 5): the Tue/Wed/Thu values are not statistically distinguishable from one another, and only the Monday-versus-Friday contrast — roughly a 9-percentage-point (≈ 28% relative) difference — is plausibly resolvable at this sample size. The full derivation, regression methodology, and cross-validation protocol appear in the unified opus (Parts III–V, hub_ops_unified_opus_v6.0.md).

κ_Z3 is the Zone-3 (PD-09–12) share of total PD-belt outbound scan volume — the PD-belt denominator, under which the three zone shares are zero-sum. It is measured on two bases that differ by a single, documented attribution rule:

The SEAS basis sits ~0.04–0.06 above the iGate basis purely because of the CCHIL → PD-09 attribution rule — not because of phase scope. A real-data check on 9 dated iGate Hub Summaries gives cumulative-final 0.328 ≈ incremental-diagnostic 0.329, confirming the two are the same flow viewed two ways. The canonical staffing source of truth is the SEAS DOW vector; the live tracker exposes both with a cchil_distribution: SEAS|iGate toggle.

This was the most consequential correction in the program. Earlier whitepaper drafts attributed scan behavior to pick-offs. Pick-offs don't scan. The correction propagated through every PPH calculation in every branch in one session.

The implication: every iGate PPH figure is loader PPH, always. Sorter and pick-off performance is invisible to iGate; it surfaces only through downstream effects — misloads (L2) and service failures (L3). This is the central fidelity proposition of Sort Intelligence: the scan stream measures one role; the operational outcome reflects three.

v3.0 reconciles the two prior SI variants into one authoritative head and applies a publishability audit. What changed: γ corrected from a stale 30-sort window value (0.958) to the full 254-sort corpus value 0.982; κ_Z3 presented as a DOW vector with a named denominator rather than a single scalar; the iGate↔SEAS duality and the CCHIL→PD-09 rule stated explicitly (both omitted before); ρ_PD/Hub corrected to a volume share, not a Pearson correlation; and statistical-confidence inflation removed throughout — sample sizes and intervals are reported, and causal language was downgraded to associational where no test supports it.

Honestly disclosed limits: per-day-of-week sample sizes are small (≈ 5–6 sorts) and the confidence intervals are reconstructed from previously reported standard errors, not re-fit from raw files; the full-year SEAS mean (0.368) and the W13–W18 window mean (≈ 0.379) are two windows of the same quantity and the ~0.011 gap is disclosed, not re-reconciled; κ_Z1 and κ_Z2 remain uncomputed (a data gap, labeled as planned work); and OJS/GEMS facts are internal-source citations.