# Watercraft: research notes Date: 2026-10-25. Scope: small craft a person paddles, pedals or rows across water: kayaks, canoes, stand-up paddleboards, inflatable boats/rafts/packrafts, and pedal-drive craft. One register for the leaf, with `craft_type` doing the dividing rather than separate registers per boat type. Out of scope, left to siblings in `supplies > active`: personal flotation devices and throw ropes (`flotation`), wetsuits/drysuits (`wetsuits`), paddles and oars sold alone (`paddles`), boards that are towed or ridden rather than self-propelled — wakeboards, kneeboards, bodyboards, surfboards, towable tubes (`watersport-boards`), snorkel/dive gear (`diving`), and pool floats (`pool-accessories`). Motorized/trailered powerboats and personal watercraft (jet skis) are also out: they are vehicles, not supplies-vertical gear, and no sibling leaf claims them either. **Overlap to flag**: stand-up paddleboards sit on the boundary with `watersport-boards`. A SUP is self-propelled with its own paddle, the same way a kayak or canoe is, so it stays here. A board with no paddle of its own (wakeboard, bodyboard, hydrofoil board) stays with `watersport-boards`. Pedal kayaks (Hobie-style) stay here as a `craft_type` value rather than becoming a new sibling, since they are still a single-paddler hull, just pedal- instead of paddle-driven. Quote labels: **verbatim** = I read the actual fetched text myself (curl plus `pandoc -f html -t plain`, or a direct JSON/API fetch) and I am quoting what that text contains. **fetched excerpt** = the fetch tool's model summarized or reworded the page for me. **search excerpt** = from a search-engine summary, page itself unread. WebSearch reported its session budget exhausted on the first three queries of this run, before any results were returned, so no claim in this register traces to a search excerpt — everything below is either verbatim-read (curl + pandoc, or a direct API fetch I read myself) or, where marked, a WebFetch model summary of a page whose raw HTML I did not also pull. ## Seed Seed run (recorded verbatim from the brief): ``` === Example entry === category: car-tyres what this shows: a number that is on every listing and still needs a judgement attached, because experts know it only compares within one manufacturer's range. [treadwear_grade] === Persona === A watchmaker, who thinks in tolerances and in what the price buys that you cannot see. ``` 1. What is watercraft's equivalent of treadwear grade? **Weight capacity.** Every kayak, canoe and SUP listing states it, and a novice reads it as a hard ceiling. Experts read it the way a tyre specialist reads treadwear: with a caveat attached. REI states the caveat directly — the number is "the total of the boat, your gear and you," not a rider-only figure — so two boats rated "400 lb" are not equally loadable if one hull itself weighs twice the other. That survived into `usable_payload_capacity` (a derived metric: capacity minus hull weight minus rider weight) and into a judgement that states the caveat plainly. 2. What would a watchmaker notice first that a typical buyer would not? Not the headline spec (length, material name) but the fit-and-finish details that decide whether the thing survives years of real use: valve quality and chamber count on an inflatable (NRS's own spec table names the exact valve model, Leafield B7 vs C7, and chamber counts separately per boat), and fabric weight (denier) rather than just the material's brand name. Kept: this pushed `air_chambers` into the facet list as a safety-relevant number rather than leaving "inflatable" as an undifferentiated blob value of `hull_material`. Dropped: a standalone `fabric_denier` facet. NRS's comparison page is the only place I found it stated at all (not on brand product pages, not on any retail listing), so the backfill cost would be high for a number almost no buyer asks for by name; folded the idea into `hull_material`'s `why_experts_care` and `backfill_note` instead of giving it its own row. ## Sources and what each supports **REI Co-op, "How to Choose a Kayak" (Ken Knapp, with A.J. Price and Lance Astles)** Archived: `http://web.archive.org/web/20261012162710/https://www.rei.com/learn/expert-advice/kayak.html` (live `rei.com` returns an HTTP/2 stream error to this session on every attempt; archive.org serves the same page). Read as raw text myself via `curl` into `kayak/rei_kayak.html` then `pandoc -f html -t plain`, so every quote below is **verbatim**. - "Where do you want to paddle? Is it a lake, a seacoast or a river?" — organizes the whole guide around `paddling_environment`. - "having a sit-in touring boat with a rudder, fixed tracking fin or a skeg (a dropdown fin) is wise" for coasts; for rivers, "a short, stable recreational sit-in or sit-on-top boat" with no rudder/skeg mentioned; "rudders are typically found on longer boats." Supports the `paddling_environment` → `tracking_aid` branch. - "Polyethylene plastic is inexpensive and abrasion-resistant, though it's the heaviest option"; "ABS... slightly more expensive... slightly lower weight... some UV resistance"; "Composites: Lightweight fiberglass and ultralight carbon-fiber boats offer a huge leap in performance and price." Supports `hull_material` values and the material trade-off. - "Weight Capacity: This is the total of the boat, your gear and you... If the boat is overloaded, it will sit too low in the water." Supports `weight_capacity`'s definition and the `usable_payload_capacity` metric. - "Longer boats cruise more efficiently... shorter hulls turn more quickly"; "Wider hulls offer more initial stability, while narrower hulls can go faster." Supports `hull_length`/`width_beam` and the fineness-ratio metric. - "Tandem kayaks... are typically more stable than their single counterparts... forgoing... the flexibility to go solo." Supports `seating_capacity`. - Pedal kayaks: "Steering is done via a rudder that's operated by a hand control... heavier than traditional kayaks... plan to trailer a pedal kayak because it is too heavy to carry on a roof rack." Supports the `pedal_boat` value of `craft_type` and keeps me from writing a "tracking_aid irrelevant on pedal boats" branch leg (it isn't — dropped). - Confidence: high. This is REI's own staff/instructor-authored buying guide, the closest thing to a specialist primer for this category. **REI Co-op, "How to Choose a Canoe"** Archived: `http://web.archive.org/web/20260210002924/https://www.rei.com/learn/expert-advice/canoe.html`. Read the same way, **verbatim**. - "Canoes in the 16 foot to 17 foot range are among the most popular." Supports `hull_length`'s typical band. - "the wider the boat, the more stable... the narrower... the more efficient." Second statement of the width trade-off (with the kayak page). - Four hull-bottom shapes (flat, rounded, shallow-arch, V) each with a named initial/secondary-stability trade. Supports `stability_profile`. - Rocker, side shape (tumblehome), freeboard, entry line — each a one-line trade. Supports `vocabulary` entries. - Materials: CrossLink3, PolyLink3 (Old Town exclusives), fiberglass, Kevlar ("about 25% lighter" than fiberglass), Royalex and Royalex Lightweight. Royalex was discontinued industry-wide in 2014 (background knowledge, not re-verified here since no canoe in today's catalog would carry it new — logged so a future author does not re-add it as a current value without checking). - "Most canoes have 2 seats, although some solo models have just 1." Supports `seating_capacity`. - Confidence: high, same basis as the kayak guide. **REI Co-op, "Paddle Board (SUP) Buying Guide" (Josette Deschambeault Rolfe, Joe Pasteris, with Gary Call and Wendell Nelson), updated Nov. 6, 2025** Archived: `http://web.archive.org/web/20260507093549/https://www.rei.com/learn/expert-advice/how-to-choose-a-stand-up-paddleboard.html`. Read **verbatim**, same method. - Hull type table: planing ("rides on top," leisure/surf/yoga), displacement ("cuts through water," fitness/touring/racing), hybrid. Not used as a facet (too deep in the weeds for the splits; the inflatable-vs-solid and length/width trades already carry the buying decision) — logged under "Trimmed detail." - "A quality inflatable SUP is designed to be inflated to 12–15 pounds per square inch." One source only; I could not confirm this figure on a second fetched page (Red Paddle Co's product page, fetched in full, does not mention PSI at all), so it does **not** go in `category.json` as a number. Kept only here. - Inflatable vs. solid table: inflatable pros (portable, cheaper, no car rack, "durable in rough conditions," softer for yoga) vs. cons ("can be less stable," "less efficient to paddle"); solid pros (best performance, custom fit, more stable, rides lower) vs. cons (storage, transport). Supports the `construction` split and a trade-off. - "Inflatable paddle boards feature PVC exteriors with drop-stitch construction." Supports `hull_material`'s `drop_stitch` value. - "Most full-sized paddle boards range from 170 to 280 liters in volume." Not used as a facet: volume is a SUP-only measure with no equivalent across the other craft types, and weight capacity (which every craft type states) already carries the same buying decision. Logged under "Trimmed detail." - "SUPs range in widths from about 25 inches to 36 inches." Supports `width_beam`'s range (with Wikipedia's kayak-width figures). - Length bands: "Short boards (under 10')... Medium boards (10' to 12')... Long boards (12'6\" and above)." Supports `hull_length`'s range. - Confidence: high, same basis. **REI Co-op, "How to Stand Up Paddleboard (SUP): The Basics"** Archived: `http://web.archive.org/web/20260519210723/https://www.rei.com/learn/expert-advice/paddleboarding.html`. Read **verbatim**. - "The U.S. Coast Guard classifies stand up paddle boards as vessels, so if you're paddling outside a surf or swimming area, you have to have a PFD on board... The Coast Guard also requires that you carry a safety whistle." Supports the judgement that SUPs (and by the same federal definition, kayaks and canoes) are regulated as vessels; corroborated directly by the regulation text itself (below). - Confidence: high. **33 CFR Part 175 (US Code of Federal Regulations, via eCFR's rendering API)**, fetched as raw HTML-tagged text and stripped with a short Python regex, so this is **verbatim** regulatory text, read by me, not a summary. Saved at `ecfr_175_api.json` in the scratch folder. - § 175.3: "Vessel includes every description of watercraft used or capable of being used as a means of transportation on the water." § 175.11: this PFD subpart "applies to all recreational vessels that are propelled or controlled by machinery, sails, oars, paddles, poles, or another vessel." Together these put every craft in this register's scope under the PFD rule. - § 175.15(a): "At least one wearable PFD is on board for each person" on a recreational vessel underway (adults need not wear it; children under 13 must, per §175.15(c)). - § 175.17(a): "Canoes and kayaks 16 feet or more in length are exempted from the requirements for carriage of the additional throwable PFD" — the extra throwable PFD, not the basic one; §175.17(b) exempts racing shells/sculls/ canoes/kayaks from any PFD requirement at all, since they are a different defined vessel class (§175.3) from a recreational canoe or kayak. - This is the regulator itself, so it is the source for the one safety judgement in this register that is a standing legal fact rather than a buying trade-off. It does not become a facet (it does not vary by product — every craft in scope is a "vessel" under it), which matches the skill's instruction to make the governing standard "a facet or a judgement." - Confidence: high. **Wikipedia, "Kayak"** (WebFetch summary of the live page; I did not separately curl the raw HTML, so this is a **fetched excerpt**, used here only to corroborate REI's claims with an independent, non-REI source, not as a sole source for any number). - "A longer kayak is faster... can also be narrower for a given displacement" but "less maneuverable." "A wide hull is more stable... shorter length. A narrow hull has less drag." Independent corroboration of REI's length/width trade-offs. - Typical lengths: touring/sea 4.9–5.8 m (16–19 ft); recreational 2.7–4.3 m (9–14 ft); whitewater under 2.4 m (8 ft), playboats 1.5–1.8 m (5–6 ft). Widths: recreational 69–91 cm (27–36 in); sea kayaks 55–65 cm (22–25 in). Two independent sources (this plus REI) now support the `hull_length` and `width_beam` ranges used in the JSON. - "Manufacturers typically produce kayaks for paddlers weighing 65–85 kg (143–187 lb), with some models for paddlers down to 50 kg (110 lb)." Single source for this specific figure (I could not confirm it on a second fetched page in the time available) — **not** used as a `range` on any facet in `category.json`; logged here only. - "Primary stability" (resistance to tipping at rest) vs. "secondary stability" (resistance to tipping once already leaning) — same pair of terms REI's canoe guide uses under "Other Design Features." Supports `stability_profile` and the vocabulary entries. - Confidence: medium (fetched-excerpt, not independently re-verified by a second raw read), used for corroboration rather than as a sole source. **Wikipedia, "Canoe"** (WebFetch summary, same caveat — **fetched excerpt**). - "The longer the waterline relative to its displacement, the faster it can be paddled"; 17 ft (5.2 m) named as popular for recreational paddlers. Second source for `hull_length`'s canoe band, alongside REI's "16 to 17 foot" figure. - "A wider boat sacrifices speed for stability." Flat-bottomed ("initial stability... sacrifice final stability and speed") vs. V-shaped ("superior speed and final stability... difficult to turn") hulls — the same shapes REI's canoe guide names, in different words, which is why I kept `stability_profile`'s values as REI's four (flat/round/shallow-arch/V) rather than Wikipedia's two. - Rocker, keel, tumblehome, freeboard — each restates a REI term in near-identical terms, which is the corroboration I used to keep those as vocabulary entries rather than drop them as single-sourced. - Sprint-canoe class rules (C1 opening length, C4 max length/weight) — racing minutiae, not used; logged under "Trimmed detail." - Confidence: medium, same caveat as the kayak article. **Wikipedia, "Inflatable boat"** (WebFetch summary — **fetched excerpt**). - "Contemporary inflatable boats are manufactured using supported fabric... rubberized synthetic fabrics, PVC and polyurethane"; Hypalon (chlorosulfonated polyethylene) named as a modern material; "some brands include inter-communicating valves that reduce the effect of a puncture." Supports `hull_material`'s `pvc`/`hypalon` values and the `air_chambers` rationale (a punctured chamber should not sink the whole boat). - Confidence: medium, used for material vocabulary only, not for any number. **Wikipedia, "Recreational Craft Directive"** (WebFetch summary — **fetched excerpt**). - The EU's four design categories (A "extended voyages," wind over force 8, waves 4 m+; B "offshore," up to force 8/4 m; C "coastal," up to force 6/2 m; D "sheltered waters," up to force 4/0.3 m) apply to "recreational craft and personal watercraft with a hull length from 2.5 metres to 24 metres," but the directive "specifically exempts canoes, kayaks, gondolas, pedalos, sailing surfboards, surfboards" from its requirements. - This is why `category.json` does **not** carry a CE design-category facet: the one safety-classification scheme that covers small recreational craft in general explicitly does not reach the craft types in this register's scope. Stated as a judgement instead, so a future author does not re-add it as a facet without re-reading this. - Confidence: medium (fetched excerpt of a Wikipedia article about the directive, not the directive's own text); the exemption clause is specific enough, and consistent enough with the US regulation's own separate racing-shell carve-out, that I am treating it as reliable for this one judgement rather than chasing the EU's own legal text. **NRS, "Compare Inflatable Raft Specs & Measurements"** `https://www.nrs.com/learn/boat-specs` — fetched live (HTTP 200), read as raw text via `curl` + `pandoc`, so this is **verbatim**. Saved at `nrs_boatspecs.html`/`.txt` in the scratch folder. - The page is itself a multi-brand spec-sheet comparison table (NRS, STAR, AIRE, Riken, Tributary — five brands across rafts, catarafts and inflatable kayaks), so it stands in for "two or three leading brands' spec pages" by design, and is the one source in this register that is a genuine `spec_sheet` rather than a primer. - Column headers read verbatim: Length, Width, Weight (lbs.), Tube Diameter, Diminishing Tube Diameter, Number of Thwarts, Bow Kick, Stern Kick, "# of Air Chambers," Valve Type, SelfBailing, Material, "Weight/Denier of Tube Material," "Weight/Denier of Floor Material," D-Rings/Handles, Number of Handles, Center Compartment, Suggested NRS Frame Width, Maximum Frame Length, Flat Area Length, Warranty, Price. Material filter options: PVC, Pennel Orca, Urethane, XR Mariner, PVC Drop-Stitch. - Inflatable-kayak rows show hull weight 57–136 lb, tube diameter 15–20 in, 4–8 air chambers, mostly PVC with 1000–2000 denier tube/floor fabric. Raft and cataraft rows show hull weight 40–179 lb, 4–8 air chambers, PVC or "Pennel Orca" (a branded urethane-coated fabric) or straight urethane. - No column for rider/total weight capacity at all — rafts are sized and compared by tube diameter and length/width, not by a stated capacity figure. This is why `weight_capacity`'s `backfill_note` says the field is commonly on the listing for kayaks/canoes/SUPs but researched-only, and often simply absent, for rafts. - The air-chamber and weight numbers above are reported here as a range taken directly from a live, multi-brand comparison of current products — not an extrapolated claim — but since it is one source, I did not put a `range.min`/`range.max` on `air_chambers` in `category.json`; the facet's `range.typical` states the band in words and this paragraph is the record for a checker to verify against the live page. - Confidence: high for what the table states; medium for how representative the five listed brands are of the whole market. **Wikipedia, "Recreational kayak"** and **Wikipedia, "Sea kayak"** (fetched 2026-10-25 for the blind-check corrections, read **verbatim** via `curl` + regex strip, saved as `wiki_Recreational_kayak.html` and `wiki_Sea_kayak.html` in the scratch folder). - "Recreational kayak" article, independent of the main "Kayak" article: "It has a wider beam of about 70–75 cm (27–30 in) for more stability in water and is generally less than 3.7 m (12 ft) in length." - "Sea kayak" article: "Solo sea kayaks are 3.0–5.5 m (10–18 ft) long, while tandem craft can be up to 8 m (26 ft) long; beam widths range from 53 cm (21 in) to 91 cm (36 in)." This is a wider span than the main "Kayak" article's sea-kayak figure (22–25 in) because it spans solo through tandem hulls, not solo alone. - Used together with the main "Kayak" article's two internally different beam figures (see below) to re-derive `width_beam`'s typical band by sub-type rather than repeat either figure uncritically. - Confidence: medium, same caveat as the other Wikipedia articles. **Re-read of Wikipedia "Kayak" for width (2026-10-25, verbatim via curl)**: the live article states recreational-kayak beam two different ways in two different sections. The "Recreational" section: "recreational kayaks have... a wider beam (69–91 cm (27–36 in)) for more stability" — this is the figure the original register used for "most hulls," and it is real text on the page today, contrary to the blind check's claim that it "does not exist on the page." But the separate "Stability" section, comparing sea and recreational kayaks directly, states "recreational kayaks, which are wider 65–75 cm (26–30 in)" — a different number for the same term, on the same page. The blind check found the second passage and concluded the first was fabricated; it exists, but the two passages disagree with each other, which is the more important finding: Wikipedia's own article on this topic is internally inconsistent about recreational-kayak width, so neither single figure should be presented alone as "most hulls." The fix below re-derives the band from all four passages (two "Kayak" sections plus the two dedicated articles) rather than trusting any one of them in isolation. **Re-parse of NRS "Compare Inflatable Raft Specs & Measurements" for air chambers (2026-10-25, live HTTP 200, `curl` into `nrs_boatspecs.html`, parsed with a short Python/regex script, not by eye)**: the page holds 13 separate per-brand tables (NRS/STAR/AIRE/Tributary/Riken rafts; NRS/STAR/AIRE/Tributary catarafts; NRS/STAR/AIRE/Tributary inflatable kayaks). Combining rafts with catarafts (152 rows with a chamber value): chambers run 3 to 9, mode 6 (62 rows, 41%), then 7 (36 rows) and 4 (34 rows); only 1 row reads 3. Inflatable kayaks alone (66 rows): chambers run 2 to 5, mode 3 (36 rows, 55%), then 4 (16 rows) and 5 (11 rows); 3 rows read 2. This confirms the blind check's finding independently (my own parse, not a re-read of theirs) and gives the exact counts behind the corrected `air_chambers.range.typical`. - Tried for a second source on the exact chamber-count numbers: Sea Eagle's own "how many air chambers" page (403), American Whitewater's raft-design wiki page (403), Wikipedia "Inflatable boat" (mentions only a 19th-century pontoon's "four different compartments," not a current product spec, and "inter-communicating valves" with no count), Wikipedia "Packraft" (a Yes/No "multi-chamber" column, no counts). None gave a usable number. WebSearch's session budget was already exhausted (same issue notes.md flagged for the original research pass). The chamber-count number stays single-sourced to NRS's own table; `category.json`'s `backfill_note` now says so explicitly instead of citing `wikipedia_inflatable_boat` as if it corroborated the figure. ## Could not reach / not used - REI's own live site: every fetch (`WebFetch` and direct `curl`) returned either HTTP 403 or an HTTP/2 "stream was not closed cleanly: INTERNAL_ERROR" from the TLS layer — bot protection, not a content issue. Used archive.org snapshots for all four REI articles instead (timestamps above), each pulled as raw HTML and read myself. - Wilderness Systems (`wildernesssystems.com/kayaks/pungo-120`) and Perception Kayaks (`perceptionkayaks.com/.../tribe-95`) both returned HTTP 200 but render their spec tables client-side in JavaScript the fetch never executes; the saved HTML has no spec numbers in it. Not used. - Red Paddle Co product page (`redpaddleco.com/products/113-sport-msl`) fetched in full (600 KB of text) and read; contains no PSI, volume or weight-capacity figures in the static HTML either. Used only to confirm the PSI figure above could not be independently checked there. - `americancanoe.org/paddlesafety` and a guessed `uscgboating.org` kayak/canoe safety page both 404. ACA's homepage redirects but I did not chase further given time; the eCFR text already gives the governing rule directly. - WebSearch: reported its session budget exhausted (200 of 200 calls used session-wide, before this task's first query returned anything) on the first attempt and every attempt after. No source in this register is a search-excerpt-only number. ## Data points used for derived metrics - `fineness_ratio` (hull_length ÷ width_beam): direction only (longer/ narrower reads as faster and tippier; shorter/wider reads as stabler and slower), from REI kayak, REI canoe and both Wikipedia pages. No numeric threshold is sourced anywhere, so the metric states the ratio and its direction and nothing more — no "above 3.5 is a touring boat" style line, because nobody sourced that line to me. - `usable_payload_capacity` (weight_capacity − hull_weight − rider weight): from REI kayak's definition of weight capacity as boat + gear + rider combined. Rider weight is supplied by the person, not a product facet. ## Against the existing add-on schema `zamana-expertise/data/schema-parts/supplies/active/add-ons/watercraft.jsonc` has `hull_material` (enum: rotomolded_polyethylene, abs, fiberglass, pvc, drop_stitch, composite, wood, hypalon, thermoform, unspecified) and a `sizing` block with `length` (ft/m/cm), `width_beam` (in/cm), `person_capacity` and `weight_capacity` (kg/lb). `active-standard.jsonc` adds `max_user_weight` (kg/lb), shared across the whole active vertical. `supplies-standard.jsonc` adds a generic `dimensions` block (length/width/height/depth, in/cm), a generic `weight` (g/kg/oz/lb) and `high_signal_markdown`. Proposals only; nothing in `zamana-expertise` was edited. | This register's facet | Existing field | Kind under AGENTS.md | Note | |---|---|---|---| | `craft_type` | none | vertical anchor | Missing entirely. The add-on has no field for "kayak vs. canoe vs. SUP vs. inflatable boat vs. pedal boat" — only `hull_material`, which conflates the craft's identity with its construction. Propose adding as a new enum; it is the fact every one of REI's four separate buying guides is organized around. | | `construction` (rigid / inflatable / folding) | derivable from `hull_material` | derived | `drop_stitch`, `pvc`, `hypalon` imply inflatable; `rotomolded_polyethylene`, `fiberglass`, `wood` imply rigid. Per AGENTS.md rule 5, a field derivable from a kept field is dead on sight — propose deriving this in the pipeline rather than adding a column. | | `hull_length` | `sizing.length` | sizing | Already present and already in the right place (the variant boundary). No change proposed beyond widening the unit enum's practical use: REI and the NRS table both give length in feet-and-inches as a single string ("9' 8\""), which the current numeric-`value` + `unit` shape handles fine once a scraper splits that string. | | `width_beam` | `sizing.width_beam` | sizing | Same as above — already correctly placed and typed. | | `seating_capacity` | `sizing.person_capacity` | sizing | Already present. Propose no change, but note the generic `active-standard.max_user_weight` field is a *different* number from `weight_capacity` below and the two are easy to conflate when writing augmentation prompts. | | `weight_capacity` | `sizing.weight_capacity` | sizing (category quality currency) | Already present. REI's definition — boat + gear + rider combined — should be the comment on this field; right now nothing in the schema says that, and `active-standard.max_user_weight` (rider only) sits one level up and could be read as a duplicate of it. Propose a one-line schema comment distinguishing the two, not a new field. | | `hull_weight` | generic `supplies-standard.weight` | sizing (reuse) | No new field needed — the generic per-listing `weight` already covers "how heavy is this object," which for a watercraft is the hull weight. Flagging only so a future augmentation pass does not duplicate it under a watercraft-specific name. | | `tracking_aid` (rudder / skeg / fixed fin / none) | none | hard filter | Missing. Buyers who read REI's coastal-vs-lake guidance filter on this directly ("I need a rudder for the bay"), and unlike most visual facts it is not reliably visible in a product photo (skegs retract, rudders fold up). Propose adding as a new enum, researched/brand-spec-page backfill. | | `air_chambers` | none | hard filter, inflatables only | Missing, and only meaningful when `hull_material`/`construction` implies inflatable. NRS's own comparison page is the only place I found it listed at all; propose adding but limiting backfill to inflatable SKUs among the top N by sales, per the skill's "lazy-load the rest" guidance. | | `stability_profile` (hull cross-section: flat / round / shallow-arch / V) | none | annotation, not a filter | Missing, and arguably should stay out of the structured schema: REI and Wikipedia both describe it in prose, no brand I fetched states it as a labeled field, and a novice cannot act on it without the paragraph of explanation the register gives it. Better candidate for `high_signal_markdown` than a new enum column. | | `paddling_environment` (calm / river / open water) | none | not a schema field | This is a buyer-intent question, not a product fact the PDP states — the closest a brand page gets is naming a use-case category ("recreational," "day touring," "touring," "whitewater") in its own marketing copy, which is `high_signal_markdown` territory, not a structured filter. Kept as a split in this register (it is how the assistant should ask), but it is not a proposal to add a schema field. | ## Branches - `craft_type` = `stand_up_paddleboard` → ask `construction`. Chain: REI's SUP guide describes construction as "inflatable vs. solid." **Revised 2026-10-25**: no longer marks `hull_material` irrelevant. The blind check caught that `hull_material`'s own value list already contains `pvc_fabric`/`drop_stitch`/`fiberglass`/`kevlar_carbon_composite` — exactly the SUP vocabulary — so it is the same facet reused, not a different sense that `construction` makes redundant. A buyer choosing a solid vs. inflatable SUP still cares whether it's carbon or fiberglass. Dropped the `makes_irrelevant` leg entirely rather than inventing an unsupported reason to keep it. - `craft_type` = `inflatable_boat_raft` → ask `air_chambers`. Chain: NRS's own comparison table is organized around exactly this craft type and lists air-chamber count as a named column for every row. SUPPORTED. **Revised 2026-10-25**: dropped `construction` from this branch's `then_ask`. The value `inflatable_boat_raft` already states its own construction; asking "rigid, inflatable, or folding?" after the buyer has already said "inflatable boat" re-asks a known answer. This also removes the structural path that let a buyer be asked `air_chambers` twice (once from this branch, once from `construction = inflatable`'s own `then_ask`, if the assistant composed both branches for the same buyer) — noted as a known overlap below rather than redesigned further, since `construction` remains an independent top-level split the assistant may still ask on its own. - `craft_type` = `kayak` → ask `tracking_aid`. `craft_type` = `canoe` → `tracking_aid` becomes irrelevant. Chain: REI's kayak guide discusses rudders, skegs and tracking fins at length for coastal/touring kayaks; REI's canoe guide, covering the same length of text on hull design, never mentions any of the three. SUPPORTED. (I considered also making `tracking_aid` irrelevant for `pedal_boat`, but REI states pedal kayaks steer "via a rudder operated by a hand control," so that leg would be false — dropped.) - `construction` = `folding` → ask `hull_weight` next (ahead of material). Chain: REI's reasons to pick a folding boat are entirely about where it lives and how it travels ("apartment... plan to travel... hike to a remote location"). **Revised 2026-10-25**: the blind check correctly caught that REI supports *why* a buyer picks folding, but never says to ask weight *before* material — that ordering is this author's own judgement, not a sourced claim. Relabeling this leg AUTHOR JUDGEMENT (reasonable, unsourced inference), not SUPPORTED. - **Known overlap, flagged 2026-10-25, not restructured**: `construction` = `inflatable` → `then_ask: ["hull_material", "air_chambers"]` and `craft_type` = `inflatable_boat_raft` → `then_ask: ["air_chambers"]` both add `air_chambers`. A raft buyer typically answers both splits (an `inflatable_boat_raft` is virtually always `construction = inflatable`), so depending on how the assistant composes branch chains across splits, it could ask `air_chambers` twice. Removing `construction` from the `inflatable_boat_raft` branch (above) removes one re-ask, but `construction` stays an independent top-level split the assistant may still walk into on its own, so the overlap on `air_chambers` itself is not structurally closed by this register — a future author or the assistant's branch-composition logic should dedupe `then_ask` targets across splits before it ships. - `construction` = `rigid` → `air_chambers` becomes irrelevant. Chain: no hardshell boat in any source read has an air chamber; the concept only exists for inflatables (NRS's table has no such column outside its inflatable sections, and REI never mentions chambers for hardshell boats). SUPPORTED (trivial but stated, since it is a real value-based exclusion). - `paddling_environment` = `open_water_coastal` → ask `tracking_aid` and `hull_length` next. Chain: REI, "a sit-in touring boat with a rudder, fixed tracking fin or a skeg... is wise" for coasts, and touring/sea boats are the longest class in both REI and Wikipedia's length bands. SUPPORTED. - `paddling_environment` = `moving_river` → `tracking_aid` becomes irrelevant; ask `stability_profile` instead. Chain: REI's river guidance ("a stable, sturdy craft that turns quickly... short recreational sit-in or sit-on-top") never mentions a rudder or skeg, and names stability and turn speed — which is what `stability_profile` encodes — as the deciding factors. SUPPORTED. - `paddling_environment` = `calm_flatwater` → `tracking_aid` becomes irrelevant. Chain: REI, "If the weather is fair... any sit-on-top or recreational sit-in boat" — no tracking aid named as needed. SUPPORTED. - `seating_capacity` = `tandem_or_more` → ask `weight_capacity` and `hull_length` next. Chain: REI kayak, tandems are "typically more stable" (a function of the wider/longer hull tandems need); REI canoe, 2-seat is the default and capacity scales with the added paddler. SUPPORTED. ## Trimmed detail - SUP hull-type table (planing/displacement/hybrid) and SUP volume in liters: real REI content, cut from the top-tier facet list because the buying decision they drive (speed vs. stability vs. maneuverability) is already carried by `construction`, `hull_length` and `width_beam`, and volume has no equivalent field on a kayak, canoe or inflatable boat's listing, which would make it a SUP-only facet in a register meant to cover the whole leaf. - Sprint-canoe racing class rules (ICF C1/C4 dimensions) from the Wikipedia canoe article: a real standard, but for a racing-class sub-market this register does not split out. - Valve type (Leafield B7/C7 and similar) and fabric denier, from the NRS table: real and safety-adjacent (a better valve is less likely to fail under pressure), but not stated anywhere outside one specialist retailer's own comparison page — folded into `hull_material`'s prose rather than given a facet of its own. - Royalex and its derivatives (canoe material, REI): discontinued since 2014, kept in `vocabulary` as background for used/legacy listings but not offered as a current `hull_material` value. ## Corrections None yet — this is the first author pass; no blind check has run against it. ## Corrections after the blind check (2026-10-25) 1. `air_chambers.range.typical` corrected from "4 to 8" to "Inflatable kayaks typically run 2 to 5 chambers, most often 3; rafts and catarafts run 3 to 9, most often 6," re-parsing NRS's live spec table myself (152 raft/cataraft rows, 66 inflatable-kayak rows; see Sources above). Confirms the checker's CONTRADICTED finding independently. 2. `width_beam.range.typical` corrected from one blended "most hulls run 27 to 36 in" line to a per-craft-type band ("Recreational kayaks run about 26 to 30 in; solo sea and touring kayaks narrow to 21 to 25 in, though tandems widen past 30 in; SUPs span 25 to 36 in"), fetched two additional Wikipedia articles (`wikipedia_recreational_kayak`, `wikipedia_sea_kayak`, added to `references`) for corroboration. Partial disagreement with the checker: the "27 to 36 in" figure does exist verbatim on the live "Kayak" article today (in its "Recreational" section), contrary to the checker's claim that it does not — but that same article states a different number (26–30 in) for the identical term in its "Stability" section, so the source is internally inconsistent rather than simply wrong. The fix addresses the underlying problem the checker correctly flagged (one blended, unqualified number) either way. 3. EU RCD judgement narrowed from "canoes, kayaks or paddleboards" to "canoes or kayaks," since `wikipedia_rcd`'s exclusion list never names paddleboards or SUPs (only "sailing surfboards, surfboards"). Re-confirmed by my own fetch of the live article. Flagging here, not fixed further: this means a SUP's status under the EU's A-to-D design categories is not actually resolved by any source in this register, and no CE-category facet exists for SUPs either, so a future author should check this before assuming SUPs are exempt. 4. Dropped `construction` from `craft_type = inflatable_boat_raft`'s `then_ask`. The value already names its own construction; re-asking it repeats a known answer. `air_chambers` stays as the branch's only `then_ask`. 5. Dropped `makes_irrelevant: ["hull_material"]` from `craft_type = stand_up_paddleboard`. The facet's own values already cover SUP's material vocabulary (`pvc_fabric`, `drop_stitch`, `fiberglass`, `kevlar_carbon_composite`); it is the same facet reused for SUPs, not a different one `construction` makes redundant, and notes.md did not actually support treating it as irrelevant. 6. Accepted and logged, not restructured: the `air_chambers` cross-branch duplication (both `construction = inflatable` and `craft_type = inflatable_boat_raft` add it to `then_ask`). Fix 4 removes one of the two re-asks of `construction` itself, but `air_chambers` can still be reached by both paths since `construction` remains an independent top-level split. See the "Known overlap" note under Branches above. 7. Added a `backfill_note` disclosing the unverified "listing" assumption to `craft_type`, `seating_capacity`, `hull_length`, `width_beam`, `weight_capacity` and `hull_weight`: all six were classified `listing` on REI's and NRS's prose, not a confirmed general-retail PDP, since every general-retailer fetch attempt (Walmart, Academy Sports, Dick's, Bass Pro) was bot-blocked for both the original author and the checker. Classifications themselves are unchanged; only the honesty of the "none"/"low" cost rating is now stated plainly. 8. Relabeled the `construction = folding → then_ask: hull_weight` ordering in the Branches trace above from SUPPORTED to AUTHOR JUDGEMENT. REI supports *why* a buyer picks folding construction but never states that weight should be asked before material; that sequencing is this register's own inference. No JSON change — the branch itself is a reasonable product decision, only its sourcing attribution was overstated. Findings not applied as JSON edits: none rejected outright. Finding 6 is addressed by documentation rather than a structural JSON change, and finding 3's "flag, don't fully resolve" element (SUP's RCD status) is likewise left open and explicitly logged rather than guessed at.