Ferris Re-Roof: Pricing Provenance

OH_RATE. 'Constants verified by read (pricing-scout, 2026-08-07)' per FERRIS-SOURCE-TABLE.md. engine/stone_pricing.py's price_from_direct_costs ships a DIFFERENT bare default (overhead_rate=0.25) and ferris_pricing.py's own call sites do not override it. quote.py explicitly passes this sourced 0.263 value for the Ferris job so net_margin/overhead_rate reproduce the reference pricing-summary.json exactly. See evidence/DISCREPANCIES.md.

NET_PROFIT_TARGET, verified-by-read 2026-08-07. engine/stone_pricing.py's bare default is 0.15; quote.py passes this sourced 0.08 explicitly for the Ferris job. See evidence/DISCREPANCIES.md.

LABOR_RATE_HOURLY, verified-by-read 2026-08-07; also independently corroborated as the Labor-table 'Crew rate' row ($65/hr per man, 2026-06-05, stone_pricing.py + constitutional rule) and as the literal $65/hr used to build both Ferris's MAN_DAY=$520 (1 man x 8hr) and Gary's crew_day=$1,040 (2 men x 8hr).

IDAHO_MATERIAL_TAX_RATE, stone_pricing.py constant per FERRIS-SOURCE-TABLE.md 'Constants verified by read' line and the Labor/Material table's 'Idaho materials tax' row (2026-06-05). Used identically by both Ferris (TAX=0.06 in ferris_pricing.py) and Gary (idaho_sales_tax: 0.06 in calculate_pricing.py rates).

MINIMUM_CHARGE, verified-by-read 2026-08-07. Appears in both Ferris's and Gary's stone_pricing 'minimums.minimum_charge' output field (not itself a binding constraint on any of the five priced Ferris options or Gary's provisional candidate, since all clear it by a wide margin).

Generic Stone Method replacement gross floor, verified-by-read 2026-08-07. FERRIS-SOURCE-TABLE.md flags this is SUPERSEDED for this job by the metal_t2 product-tier doctrine (floor 0.60) under 'Margin doctrine (GOVERNING for this job)'. Not applied to any Ferris option; carried for completeness only.

Generic Stone Method replacement gross target, verified-by-read 2026-08-07. Superseded for Ferris by the metal_t2 doctrine target range 0.65-0.70 (see metal_t2_gm_floor / metal_t2_gm_target_range entries). Not applied to any Ferris option; carried for completeness only.

difficulty_pricing.py PRODUCT_GM_DOCTRINE (mitch-directed 2026-07-21, Jones run); restated in 2026-08-07 Mac handoff README. Skill copy and handoff package copy verified byte-identical 2026-08-07 per FERRIS-SOURCE-TABLE.md.

PRODUCT_GM_DOCTRINE['metal_t2']['floor'] in engine/difficulty_pricing.py (byte-identical to the FERRIS-SOURCE-TABLE.md-cited copy per its freshness audit). Read via gm_bounds_for_product('metal_t2', tier) as MAIN_FLOOR / CAB_FLOOR in ferris_pricing.py.

PRODUCT_GM_DOCTRINE['metal_t2']['target_range'] in engine/difficulty_pricing.py. Position inside the range is set by job difficulty tier via gm_bounds_for_product: main house scores T2_standard -> target 0.65 + (0.70-0.65)*(1/3) = 0.666667; cabins score T1_simple -> target 0.65 (bottom of range).

Main house: 6:12 predominant (+10), 1 story (+5), 6 facets (+10), interfaces 19'1" step (+3) + 1 chimney (+4) + 2 pipe jacks (+2) + 46'5" valley (+3) = 12 -> score 37 -> tier T2_standard -> GM target 0.6667. Inputs from Roofr 2026-08-05 + CompanyCam photos, verdict KNOW.

Cabins: 6:12 (+10), 1 story (+5), 2 facets each (+5), no interfaces (0) -> score 20 -> tier T1_simple -> GM target 0.65 (bottom of range). Source: GAF QuickMeasure 5437438, verdict KNOW.

22ga carries +0.015 GM over the 24ga target ('heavier gauge never sells thinner'). FERRIS-SOURCE-TABLE.md verdict is explicitly 'session-decided, flagged' (a policy call made in-session, not a dated external document) -- tagged PROXY here because it is a derived/assigned figure, not read off a priced sheet.

ASC Design Span 16in 24G DuraTech XL (PVDF), $4.99/LF ($374.24/SQ), from pricing-data-extracted.md (ASC/ABC extract). Tagged PROXY (per task instruction) because AEP Span-Lok hp itself is quote-only (web lane, 2026-08-07); ASC Design Span is 'the nearest priced cousin' (same ASC Profiles family), not a direct Span-Lok hp quote. Source table's own verdict column says CURRENT-but-aging (about the ASC sheet's freshness); this table elevates it to PROXY because of the product substitution.

22ga panel = 24ga $4.99/LF x 1.20 = $5.99/LF. No 22ga dollar price exists locally or publicly. Ratio basis: McElroy Metal (manufacturer) 2020-10-13, direction confirmed Sheffield Metals 2025-02-21; ratio itself fetched 2026-08-07. Verdict DERIVED-PROXY in source table.

Ibex ML clips, reroof-patterns.md. Verdict CURRENT.

Ibex, reroof-patterns.md. Verdict CURRENT.

Collins budget planner. Verdict CURRENT-but-aging ('only dated figure'); still a dated first-party sheet, so tagged VERIFIED rather than PROXY.

Malarkey 401 SA HT class, reroof-patterns.md; ABC Hayden catalog $48.50/SQ agrees (2026-03-05). Verdict CURRENT.

Titanium UDL-30 class over nail-base, ABC Hayden catalog. Verdict CURRENT-but-aging; dated first-party sheet, tagged VERIFIED.

Vented nail-base polyiso, 3in foam + vent + nailing surface. THE LARGEST PROXY in the job (task-flagged). DERIVED: foam-only proxy $0.72/sqft-inch x 3in = $2.16 (Planner5D/RoofCalc 2024-2025 guides) + wood nailing surface $1.31/sqft (Collins plywood $42/sheet, 2026-03-17) + factory vent-spacer/assembly premium $0.53 = $4.00/sqft. Product is quote-only everywhere public (QXO/Dealers Choice login-gated, verified 2026-08-07). LOUD FLAG per source table and per pricing-proof.md.

Heavy through-insulation fasteners. Allowance; Atlas pattern 30+/panel is design output, but the unit PRICE is DERIVED-PROXY: 'no dated heavy-fastener row exists; Ibex 1in screws $0.077 too short to substitute.' Atlas doc itself is undated; 2026-08-05 is the fetch/index date, not a price date.

Ridge cap, custom insulation-depth profile. Derived: Ibex base $24.05/pc (2026-04-16, reroof-patterns.md) x 1.5 custom-depth uplift = $36.08. Uplift basis: 'Campos custom precedent x1.05 was gauge-like-for-like; deeper profiles need more break time' -- the 1.5x multiplier itself is not independently dated, flagged in source table.

Eave/rake drip edge, custom depth. Derived: Ibex base $7.66/pc (2026-04-16) x 1.5 custom-depth uplift (same basis/flag as ridge_trim_custom_depth) = $11.49.

ABC Hayden catalog (Award 26ga; 24ga fabrication uplift inside trim uplift). Verdict CURRENT-but-aging; dated first-party sheet, tagged VERIFIED.

Ibex, reroof-patterns.md. Verdict CURRENT.

Seam-clamped snow retention (S-5 ColorGard class, materials only). S-5 manufacturer knowledge-base materials-only range $4.51-$13.79/LF; midpoint carried pending engineered layout. Verdict DERIVED-PROXY (engineering-pending), flagged.

Dektite HT class, ABC Hayden catalog. Verdict CURRENT-but-aging; dated first-party sheet, tagged VERIFIED.

Ibex, reroof-patterns.md. Verdict CURRENT.

Williams/Collins evidence, pricing-data-extracted.md. Verdict CURRENT-but-aging; dated first-party sheet, tagged VERIFIED.

pricing-data-extracted.md. Verdict CURRENT-but-aging; tagged VERIFIED.

reference_material_pricing.md (Renee Pollard email 2026-03-23), effective 2026-04-10. NOT APPLIED -- no Metal America 26/29ga rows used in this build. Carried for completeness only ('every other line the source table carries').

Schmidt-lane freshness intel relayed by lead, 2026-08-07; effective 2026-06-15. Supplier-specific; NOT APPLIED -- no Metal America rows used in this build. Carried as a MARKET-MOVED flag on the ASC 24ga panel row (the current ASC price is unverified and could be similarly higher); a dated-basis flag, not an applied escalation.

$65/hr per man, 2-man crew context; MAN_DAY constant in ferris_pricing.py is the PER-PERSON day rate ($65 x 8hr = $520); 'man_days' quantities on each option (30.5 / 32.5 / 8.0 / 8.5 / 1.0) are total person-days, not calendar days. Source: stone_pricing.py + constitutional rule, 2026-06-05.

Williams/Collins planner man-day figures. Verdict SUPERSEDED by the $65/hr crew rate above -- DEAD, NOT used in current pricing. Listed only because the source table carries it; source_date is NOT_FOUND because the table gives no date for this dead figure (its row reads '-').

reroof-patterns.md. Verdict CURRENT. Informational: ferris_pricing.py does not apply this as an explicit multiplier to a snap-lock baseline -- the 30.5/32.5/8.0/8.5/1.0 man-day figures are hand-set per-job values anchored directly to the Collins v22.5 man-day precedent below, which already reflects mechanical-seam + nail-base labor intensity.

ENGINEER-v22.5-pricing.md. Verdict CURRENT (newest real ledger). Ferris labor model (30.5 md 24ga main / 32.5 md 22ga / 8 md cabins / 1 md counterflashing at $520/man-day) is anchored to this figure, adding the nail-base layer + mechanical seam on top per pricing-proof.md Assumptions.

No written same-coil provenance exists (repo Source-Index 2026-08-05; Gmail re-searched by the Mac 2026-08-05). Stays unpriced, presented as a procurement gate on the Ferris job. Presenting it priced would violate the repo Pricing-Status rule. Also carried unpriced in Ferris's own pricing-summary.json unpriced_gates block.

T&M per terms, not a bid line. FERRIS-SOURCE-TABLE.md gives no independent date for this specific rate; source_date uses the table's own generation date. Cross-corroborated verbatim by Gary's copied calculate_pricing.py/pricing-summary.json 'unsuitable_decking_per_4x8_sheet': 225 (2026-08-05). Ultimate origin is Ketron's canonical terms.md pre-approved change-order list, which is not in this build's copied source set.

Fascia $15-25/LF, pre-approved change-order rate, T&M per terms. Same dating caveat as deck_repairs_decking_per_sheet. Cross-corroborated verbatim by Gary's copied calculate_pricing.py 'fascia_repair_per_lf_range': [15, 25].

'The $185/layer removal rate was dropped 2026-08-07: it is not in canonical terms.md's pre-approved list, and a signed document must not carry two different pre-approved lists.' NOT usable for Ferris; status NOT_FOUND reflects that Ferris has no valid current source for this rate, even though the number is visible in the doc. CONTRAST: Gary's copied calculate_pricing.py still carries this exact figure live as 'additional_shingle_layer_per_square': 185 -- see gary_additional_shingle_layer_per_square below. This is a genuine cross-job inconsistency in the source material, not a transcription error here; noted in evidence/DISCREPANCIES.md.