# Facial Tissues, Paper Napkins & Paper Towels: research notes

Date: 2026-10-25. Scope: the three disposable paper products the taxonomy leaf names in its label — facial tissue (box or pocket tissues for the nose and face), paper napkins (tissue or airlaid, beverage/luncheon/dinner), and paper towels (kitchen roll). One register, three branches off a `product_type` split, not three registers. Explicitly out of scope and left to sibling leaves in `catalog.json` (same `supplies > household` prefix): toilet/bathroom tissue and flushable wipes (`toilet-paper-wipes`), the holder/dispenser hardware these products sit in (`dispensers`), and reusable cleaning cloths, sponges and scouring pads (`cloth-pad-sponges`). Also out of scope: cloth table napkins (textile, not this leaf, not in the household group at all) and disposable "guest towels" / airlaid hand towels for parties, which I found on a napkin manufacturer's own site filed as a napkin-adjacent SKU but which function as bathroom hand towels, not kitchen or table paper — flagging rather than swallowing, since no sibling leaf obviously owns them either.

Quote labels: **verbatim** = I fetched the raw page myself with curl and read the converted text (or curl's text/pandoc output directly), or WebFetch returned the quoted sentence inside quotation marks matching what I could independently verify from the same page's raw structure; **fetched excerpt** = the WebFetch tool summarised or lightly reworded the page and I did not separately pull the raw text; **search excerpt** = would have come from a search-engine summary with no page read — WebSearch's budget was exhausted for this session before I could run any queries (every query returned "session has used its web search budget"), so there are no search-excerpt sources in this run at all; everything below is either verbatim (raw curl + pandoc, the overwhelming majority) or fetched excerpt (WebFetch only, logged where used).

Tool note to report: WebFetch cannot reach `web.archive.org` at all ("Claude Code is unable to fetch from web.archive.org") — every archive.org page had to go through curl instead, and curl itself got rate-limited once ("suspected abusive bot traffic") on back-to-back wayback requests, recovering after a ~20s pause. Recorded as requested; not a category-content issue.

## Seed

Seed run output (recorded verbatim from the brief):

- Example: bedsheets `weave` — "the expert split is often not the number everyone advertises. Thread count is the generic trap here. Weave is what changes how the sheet feels and sleeps."
- Persona: "A print production manager, who thinks in what the spec sheet says versus what comes off the line."

1. **The paper-tissue-towels equivalent of weave is ply plus basis weight (GSM), not sheet count.** Sheet count is the number every pack shouts ("308 sheets!"), and it is the thread-count-style generic trap: it is real, it is on every listing, and it tells you almost nothing about how much paper you are actually getting once sheet length varies ("select-a-size" halves the sheet without halving the count the same way). Ply (how many of the base tissue layers are stacked) and basis weight (how heavy each layer is, in g/m²) are what an expert actually reads, the same way weave beats thread count for sheets. Both survived the data check: ply is already a dedicated field in the existing add-on schema, and two independent Wikipedia articles (tissue_paper, Yankee_dryer) give consistent GSM bands for facial tissue, towels and the tissue-paper family generally. Kept as the `ply` split and the `basis_weight_gsm` annotation facet.
2. **The print production manager's instinct — spec sheet versus what comes off the line — maps onto fibre source.** A recycled or bamboo roll's spec sheet (sheet count, size, ply) can read identically to a virgin-pulp roll's, but what comes off the line is different: recycled fibre has already been shortened by an earlier papermaking cycle, so the sheet needs a resin or extra virgin fibre stitched back in to hold together, which a buyer never sees stated anywhere. This is exactly the production-manager gap between the spec and the reality. Kept as the `fiber_source` split, with `wet_strength_treated` as the facet that actually answers whether the gap was closed.

Dropped from the seed's first instinct: softness itself as a standalone split. It is real (debonding agents raise softness and lower strength — see madehow below) but it is not independently answerable by a novice and not stated on listings; it rides along inside `ply` and `fiber_source` instead, and the softness-vs-strength trade-off and visual grammar below carry it.

## Sources and what each supports

### Primers: how the paper itself is made

**Wikipedia, Facial tissue**
https://en.wikipedia.org/wiki/Facial_tissue
- verbatim (curl + pandoc): "Facial tissue and paper handkerchiefs are made from the lowest basis weights tissue paper (14–18 g/m²)... These paper types consist usually of 2–3 plies... the base tissue is normally made entirely from pure chemical pulp, but might contain added selected recycled fiber... The tissue paper might be treated with softeners, lotions or added perfume... Facial tissue may contain non-biodegradable additives for strength."
- verbatim: "Kimberly-Clark also introduced pop-up, colored, printed, pocket, and 3-ply facial tissues."
- Supports: facial-tissue GSM band and ply count, lotion/softener treatment, the strength-additive fact. Confidence high.

**Wikipedia, Tissue paper**
https://en.wikipedia.org/wiki/Tissue_paper
- verbatim: "Tissue paper, particularly low-density tissue paper, is produced on a paper machine that has a single large steam heated drying cylinder (Yankee dryer)... Creping is done by the Yankee's doctor blade that is scraping the dry paper off the cylinder surface."
- verbatim: "The highest water absorbing applications are produced with a through air drying (TAD) process. These papers contain high amounts of NBSK and CTMP. This gives a bulky paper with high wet tensile strength and good water holding capacity. The TAD process uses about twice the energy compared with conventional drying of paper."
- verbatim: "Paper towels are the second largest application for tissue paper in the consumer sector. This type of paper has usually a basis weight of 20 to 24 g/m². Normally such paper towels are two-ply... Normally, some long fibre chemical pulp is included to improve strength."
- verbatim: "Table napkins can be made of tissue paper. These are made from one up to four plies and in a variety of qualities, sizes, folds, colours and patterns depending on intended use and prevailing fashions."
- Supports: the creping mechanism, TAD vs conventional drying and its energy cost, paper-towel GSM and ply, napkin ply range (1–4), long-fibre-for-strength fact. Confidence high.

**Wikipedia, Yankee dryer**
https://en.wikipedia.org/wiki/Yankee_dryer
- verbatim: "A Yankee dryer is a pressure vessel used in the production of machine glazed (MG) and tissue paper. On the Yankee dryer, the paper goes from approximately 42–45% dryness to just over 89% dryness."
- verbatim: "Creping is done by the Yankee's doctor blade scraping the dry paper off the cylinder surface thereby crêping the paper."
- verbatim: "This is due to the necessity of creping and made possible by the low grammage (gsm = gram per square meter) of the paper sheet for tissue products, which is in the range of 14-45 gsm. For the production of the higher gsm in this range, some machines are nevertheless provided with some (4-10) drying cylinders after the yankee. In this case one speaks of 'wet creping'."
- Supports: second, independent source for the GSM band across the whole tissue-product family (14–45 gsm), consistent with facial tissue at the low end and heavier napkins at the high end; the mechanism (creping) itself. Confidence high.

**Made How, "Tissue with Lotion"** (covers facial tissue; explicitly names paper towels, napkins and toilet tissue as the same product family)
https://www.madehow.com/Volume-3/Tissue-with-Lotion.html
- verbatim: "Facial tissues belong to a class of paper products used extensively for personal hygiene in modern society. Other products of this type include paper towels, napkins, and sanitary (or toilet) tissue."
- verbatim: "softness can be increased by adding agents that interfere with the way the fibers within the tissue interact, making them less closely bonded to each other. These are known as debonding agents. However, these materials tend to decrease the tensile strength of the fabric and may irritate skin on contact."
- verbatim: "Enhanced softness can also be achieved by coating the fabric with oily materials. However, this limits the amount of moisture the tissue can absorb. In fact, the coating can also make the fabric so hydrophobic (water hating) that it can not be processed properly in sewage plants."
- verbatim: "a pattern may be imprinted in the tissue by juxtaposing the web on an array of supports during the dewatering process... These are regions of varying fiber density and are visible as tiny patterned 'pillows' on the final sheet" (embossing mechanism).
- verbatim, lotion dosing: "Ideally low amounts of lotion are applied (0.3% or less) to prevent the web from being overcoated."
- verbatim, testing methods: softness is "assessed by subjective evaluation by trained panelists... paired comparison, as described by the American Society for Testing Materials (ASTM)... reported in Panel Score Units"; absorbency is "wetting time" measured by crumpling tissue into a ball and timing how long it takes to saturate in a beaker of water; linting is measured with "a motor driven device known as the Sutherland Rub Tester."
- Supports: the debonding-agent softness/strength trade-off, the lotion/sewage trade-off, the embossing ("pillow") mechanism, and that softness and absorbency are measured by lab panel/wetting-time tests that never reach a retail listing. Confidence high.

### Fibre source

**Wikipedia, Recycled paper**
https://en.wikipedia.org/wiki/Recycled_paper
- verbatim: "After repeated processing the fibres become too short for the production of new paper, which is why virgin fibre (from sustainably farmed trees) is frequently added to the pulp recipe."
- verbatim: "Three categories of paper can be used as feedstocks for making recycled paper: mill broke, pre-consumer waste, and post-consumer waste... The industrial process of removing printing ink from paper fibres of recycled paper to make deinked pulp is called deinking."
- Supports: the core recycled-fibre weakness fact and the deinking vocabulary. Confidence high.

**Wikipedia, Bamboo** (the dedicated "Bamboo paper" article redirects here; section "Paper")
https://en.wikipedia.org/wiki/Bamboo
- verbatim: "The average fiber length is similar to hardwoods, but the properties of bamboo pulp are closer to softwood pulps due to it having a very broad fiber length distribution."
- verbatim: "Bamboo is cut after three years of its germination. No cutting is done during the rainy season (July–September)."
- verbatim (growth rate, general bamboo biology, not paper-specific): "Certain species of bamboo can grow 91 centimeters (36 inches) within a 24-hour period... Growth up to 120 centimeters (47.2 in) in 24 hours has been observed in the instance of Japanese giant timber bamboo."
- Single source. I looked for a second (a dedicated "Bamboo paper" or "Bamboo pulp" Wikipedia article does not exist — both redirect to the general Bamboo article; brand sustainability pages for bamboo tissue brands either blocked the fetch or redirected without reaching content, logged under "Could not reach"). The fibre-property and 3-year-harvest claims go in the JSON labelled as resting on this one source.
- Supports: bamboo pulp's strength character despite the fast growth cycle, and the harvest-age fact. Confidence medium (single source, flagged).

**Marcal, "Our Story"** (read via a 2017 Internet Archive snapshot; the live marcalpaper.com blocked the fetch with a 403)
http://web.archive.org/web/20170606035402/http://marcalpaper.com/our-story
- verbatim: "For over five decades, Marcal has been reducing landfill by making our paper products from recycled paper. We use paper collected from residential neighborhoods, from the small blue baskets in office buildings, from unwanted junk mail, and waste from printers."
- Supports: that a real, long-running mainstream brand runs on 100% recycled fibre (existence proof for the `fiber_source: recycled` value, not a performance number). Confidence high for what the brand states; this is the brand's own description, not an independent test.

### A real brand's own account of what matters: Bounty

**Wikipedia, Bounty (brand)**
https://en.wikipedia.org/wiki/Bounty_(brand)
- verbatim: "While most paper towels were being marketed promoting their strength or softness, P&G found consumers primarily preferred absorbency... Bounty replaced Charmin towels in 1965, and introduced a new 2-ply towel."
- verbatim: "In 1982, P&G Senior Inventor Paul Trokhan developed a proprietary technology that enhanced Bounty's absorbency by creating micro-regions within the towel focused on different attributes like strength and softness."
- verbatim: "Consumer Reports reported (2014) the best paper towel was Bounty DuraTowel, followed by the next two on the list also being Bounty products." "In 1998, Bounty started selling napkins."
- Supports: that absorbency, not strength or softness alone, is what towel buyers actually optimise for (one company's market research, stated as such); that the embossed "micro-region" pattern — the same imprinting mechanism madehow describes — is the patented lever for balancing the three. Confidence medium: this is P&G's own framing of its own research, repeated by Wikipedia, not an independent study, and I say so in the judgement.

### Napkin sizes (the one place I leaned on a manufacturer-retailer's own product pages)

**Napkins.com (Hoffmaster), three product pages, each fetched and read directly**
- White 2-Ply Beverage Napkins: https://www.napkins.com/products/white-2-ply-beverage-napkins — verbatim: "Size: 5" x 5"", "Material: 2-Ply Tissue", description "each 2-ply napkin measures 4.938 inches x 4.938 inches while folded."
- Navy 2-Ply Luncheon Napkins: https://www.napkins.com/products/navy-2-ply-luncheon-napkins — verbatim: "Size: 6.5" x 6.5"", description "measures 6.375 inches x 6.375 inches" folded.
- Wholesale 8.5x8.5 Regal Embossed White Dinner Napkins: https://www.napkins.com/products/wholesale-8-5-x-8-5-regal-embossed-white-dinner-napkins-2000-ct-hf-180705 — verbatim: "Size:" 8.5", description "these high-quality 3-ply napkins measure 8.5"."
- Supports: the beverage/cocktail (5"×5"), luncheon (6.5"×6.5") and dinner (8.5"×8.5") folded-size ladder, each exactly double a quarter of 10, 13 and 17 inches — consistent with the standard foodservice quarter-fold convention, though I could not reach a second, independent source (a trade body or a second manufacturer) stating that convention in words; logged under "Could not reach." Confidence medium-high: internally consistent across three separately fetched pages from the same manufacturer family, but a single source family.
- Also from this site: White 3-Ply Guest Towels ("Made from Premium 3-Ply Recycled Tissue," 4.375" × 8") — the disposable-hand-towel product flagged as out-of-scope in the scope line above, kept here as the source for that flag.
- Also: product names on this site reliably distinguish "linen-like" (airlaid-style, cloth-feel) napkins from plain "tissue" napkins in the title itself, e.g. "14 x 14 Linen-Like Natural White Dinner Napkins" versus "White 2-Ply Beverage Napkins" — supports `napkin_material_format` being listing-sourced (title), not researched.

### Certification and food contact

**Wikipedia, Forest Stewardship Council**
https://en.wikipedia.org/wiki/Forest_Stewardship_Council
- verbatim: "The FSC Chain of Custody (CoC) system allows the tracking of FSC certified material from the forest to the consumer... Only companies that have FSC chain of custody certification are allowed to use the FSC trademarks and labels."
- verbatim: "FSC-Watch commonly accuses FSC of practices that are a form of greenwashing." "In 2018, Greenpeace International chose not to renew its membership with FSC, claiming that FSC is not meeting its goals of protecting forests and human rights."
- Supports: what the FSC label actually certifies (chain of custody, not a performance claim about the sheet itself) and that its credibility is genuinely disputed by named former backers, not a settled fact. Confidence high for what each side said; the underlying dispute itself is unsettled, which the judgement states.

**21 CFR 176.170 and 176.180, via Cornell Legal Information Institute**
https://www.law.cornell.edu/cfr/text/21/176.170 and https://www.law.cornell.edu/cfr/text/21/176.180
- verbatim, 176.170: "Substances identified in this section may be safely used as components of the uncoated or coated food-contact surface of paper and paperboard intended for use in producing, manufacturing, packaging, processing, preparing, treating, packing, transporting, or holding aqueous and fatty foods."
- Supports: that any napkin or paper towel sold in the US for use with food already has to meet a federal food-contact-substance rule before it reaches a shelf, so there is no further consumer-facing safety certification to check for ordinary use. Confidence high; this is the primary regulation itself, which is why it stands as a single source for a scope statement rather than a disputed number.

### Could not reach

- EPA's Comprehensive Procurement Guideline recycled-content recommendations for sanitary tissue products: the modern epa.gov URLs for this guideline return 404 (the page has moved or been retired since the last run that might have cited it); I could not find the live replacement without a working search tool. No recycled-content percentage threshold is in the JSON as a result.
- NRDC "The Issue with Tissue" report (nrdc.org): 403 on the live fetch; no wayback snapshot resolved either. Would have been the main source for a bamboo-vs-recycled-vs-virgin forest-impact comparison with numbers; without it, the register's sustainability framing stays qualitative (fibre mechanics only), not quantified in hectares or gallons.
- FSC.org itself (faq-page, forest-products/pulp-and-paper): 403 and 404. The FSC side of the dispute above comes from Wikipedia's account of FSC's own published materials, not FSC's site directly.
- Tork/Essity's napkin and towel product pages (tork.us.com, tork.co.uk): the US domain would not resolve for curl at all; the UK page loaded but is a JavaScript shell with no text content in the raw HTML. Would have been a commercial spec-sheet source for GSM and embossing terminology on real SKUs.
- Bounty's, Viva's and Puffs's own sites (bounty.com, vivapapertowels.com, puffs.com): bounty.com loads but is a JavaScript-rendered marketing shell with no technical content in the raw HTML; vivapapertowels.com would not resolve for curl; puffs.com would not resolve. No wayback snapshots resolved for the specific pages I tried (types-of-paper-towels, select-a-size). "Select-a-size" therefore stays out of the JSON as a named, cited feature; the register instead treats sheet length/width as part of the existing `sizing.dimensions` block, which already covers it without a new field.
- A second, independent source for the foodservice napkin quarter-fold size convention (10/13/17 inch unfolded, 5/6.5/8.5 folded): not found; the three napkins.com pages above are one source family. Flagged as a single-source number.
- A second source for the bamboo-pulp fibre-property and three-year-harvest claims: not found (see Fibre source above).
- Webstaurantstore's napkin buying guide: 403 live, no wayback snapshot resolved.
- WebSearch itself: exhausted for the entire session before a single query ran ("this session has used its web search budget (200 of 200 WebSearch calls)"). Every source below came from a direct fetch of a URL I already knew or found by following links inside a page I had already fetched (Wikipedia's own cross-links, mostly), never from a search result.

## Data points used for derived metrics

- `cost_per_100_sheets`: pure arithmetic from price and `sizing.sheet_count_per_unit` (both already in the existing schema). No external source needed for the formula itself; it exists because sheet count is the number brands lead with and the one that becomes misleading once sheet length varies (see the seed answer above).
- `total_sheet_area_in2`: pure arithmetic from `sizing.dimensions.length`, `sizing.dimensions.width` and `sizing.sheet_count_per_unit`. Exists for the same reason: a "select-a-size" or half-sheet roll can show a higher sheet count than a full-sheet roll while holding less paper overall.

## Against the existing add-on schema

Add-on read: `supplies/household/add-ons/paper_tissue_towel.jsonc` — currently just `ply` (number) and a local `sizing` block with `sheet_count_per_unit` and `roll_count`. Composed with `household-standard.jsonc` (`materials` — includes `paper` and `cardboard_paperboard`, usable as-is; `mounting_install` — not applicable to loose sheets, noise for this add-on; `sizing.weight_capacity` — noise, nothing in this leaf is load-bearing) and `supplies-standard.jsonc` (`high_signal_markdown`; shared `sizing` with `dimensions`, `capacity`, `weight`, `pack_count`). Field-kind vocabulary (AGENTS.md): vertical anchor, quality currency, hard filter, sizing, high-signal markdown.

| Facet in this register | In add-on today | Proposed kind (AGENTS.md) | Note |
|---|---|---|---|
| `product_type` | **no field at all.** The whole leaf has one add-on with no way to tell a facial tissue listing from a napkin listing from a towel listing except by parsing the title. | hard filter | The single biggest gap found. Everything downstream (lotion, napkin size, wet strength) depends on knowing which of the three a row is. |
| `ply` | yes, already present | hard filter / quality currency | Fits as-is. Typical ranges differ enough by `product_type` (facial 2–3, towel usually 2, napkin 1–4 — Wikipedia tissue_paper, Yankee_dryer) that it is worth noting in a comment rather than changing the field. |
| `fiber_source` | no | hard filter | People search "recycled paper towels" and "bamboo napkins" by name; the word is reliably in the title (Marcal, napkins.com "linen-like" lines) so this is cheap to extract, not research-only. |
| `lotion_additive` | no | hard filter (boolean) | Facial-tissue-specific; "Lotion," "Ultra Soft," "Anti-Viral" are variant names on the box and the title. |
| `scent_dye_free` | no | hard filter (boolean) | "Unscented," "fragrance-free," "dye-free" are bullet/title claims across all three products, not just facial tissue. |
| `napkin_size` | no field; **derivable from `sizing.dimensions`**, which already exists in supplies-standard | derived | Once `dimensions.length`/`width` are populated for a napkin row, beverage (~5"/10"), luncheon (~6.5"/13") and dinner (~8.5"/17") fall out of the convention bands sourced above. No new field needed — the gap is that `dimensions` is not being populated for single-sheet products today. |
| `napkin_material_format` | no | hard filter | "Linen-like" vs plain is in the title (napkins.com). |
| `basis_weight_gsm` | no | quality currency, but an honest one | This is the real comparable number (the thread-count-for-sheets equivalent), but it is a trade/spec-sheet number, not a retail-listing one; I would not commit to augmenting it on every row at low cost. Mark researched, medium–high backfill cost, best-effort on top brands only. |
| `wet_strength_treated` | no | **not a reliable field; high-signal markdown instead** | Presence of a "stays strong when wet" bullet signals yes; absence does not reliably mean no. Does not clear the AGENTS.md bar for a hard filter ("must exist on every row"). |
| `embossing_pattern` | no | **visual, not a schema field** | This is read off the photo (smooth vs quilted vs printed), which is exactly the AGENTS.md rule-1 case: "if a field can be better determined by looking at the product photo than by reading the page, it doesn't belong here." Kept in this register as an annotation/visual-grammar fact, flagged here as image-embedding territory, not a text field to add. |

Nothing in the existing add-on is noise by the AGENTS.md test; it is simply thin (2 fields for a 3-product leaf).

## Branches

- `product_type` = `facial_tissue` → then_ask `lotion_additive`, `scent_dye_free`; makes_irrelevant `napkin_size`, `napkin_material_format`, `wet_strength_treated`. Chain: lotion/softener treatment is a fact stated specifically about facial tissue and paper handkerchiefs (Wikipedia facial_tissue, madehow); napkin size and material-format are napkin-only facts by definition; wet strength is the paper-towel/table fact (next leg) that a tissue used on the face or nose has no stated reason to need. SUPPORTED for lotion/scent and for the napkin facts (definitional); the wet-strength exclusion is reasoned, not directly stated, and is the weakest leg here — kept because the alternative (asking a facial-tissue buyer about wet strength) would be the generic-trap failure the checks.md log warns about.
- `product_type` = `paper_napkin` → then_ask `napkin_size`, `napkin_material_format`, `wet_strength_treated`; makes_irrelevant `lotion_additive`. Chain: napkin size and material-format are the napkin-specific facets by definition; lotion/softener coating is stated specifically about facial tissue (Wikipedia facial_tissue, madehow), never about napkins. SUPPORTED.
- `product_type` = `paper_towel` → then_ask `wet_strength_treated`, `basis_weight_gsm`; makes_irrelevant `lotion_additive`, `napkin_size`, `napkin_material_format`. Chain: "Resin size is used to improve the wet strength" is stated specifically in the context of paper towels (Wikipedia Paper_towel, in the general paper-towel entry); lotion and napkin facets are definitionally the other two branches'. SUPPORTED for wet strength; napkin/lotion exclusions definitional.
- `fiber_source` = `recycled` → then_ask `wet_strength_treated`. Chain: recycled fibre is shortened by earlier processing and "virgin fibre... is frequently added" to compensate (Wikipedia Recycled_paper, verbatim) — whether that compensation happened is exactly what `wet_strength_treated` asks. SUPPORTED, single stated sentence.
- `fiber_source` = `bamboo` → makes_irrelevant `wet_strength_treated`. Chain (two stated facts): bamboo pulp's "properties... are closer to softwood pulps" (Wikipedia Bamboo, verbatim) and paper towels specifically get "long fibre chemical pulp... included to improve strength" where that long fibre is the softwood-type fibre (Wikipedia Tissue_paper, verbatim, same strength-via-long-fibre logic). Chaining these: bamboo pulp already behaves like the strength-giving long fibre that resin/wet-strength treatment exists to compensate for its absence, so the question matters less. WEAKER than the other legs (it is an inference across two sentences about two different things — general pulp science and paper-towel fibre blending — not one sentence about bamboo paper towels specifically); kept because both halves are individually verbatim and the inference is the ordinary papermaking logic the rest of the register relies on, but flagged here in case a blind check wants it cut.
- `ply` → no branches. No sourced fact ties a specific ply count to a downstream facet becoming more or less relevant (ply and wet-strength treatment are independent levers per the sources above); an untraceable branch here would be exactly the "branches keyed to values the question cannot support" failure mode in checks.md, so none was added.

## Trimmed detail

- **mechanism.plain**: cut "a step called creping" gloss (kept in vocabulary); cut the exact Yankee-dryer dryness figures (42–45% to 89% dryness) and the "4–10 additional cylinders... wet creping" detail for heavier grades (kept in the Yankee dryer source entry above); cut the ASTM Panel Score Units and Sutherland Rub Tester testing-method detail entirely (kept in the madehow source entry and in the `basis_weight_gsm`/judgements below as the reason lab numbers never reach a listing).
- **summary**: cut "a resin coating" detail for wet-strength and the specific "0.3% or less" lotion dosing figure (kept in the madehow source entry and in `lotion_additive`'s note).
- **fiber_source.why_experts_care**: cut the specific "broad fiber length distribution" phrase for bamboo and the three-year harvest-age fact (kept in the Fibre source source entry and in the bamboo judgement below); cut the deinking vocabulary detail (kept in vocabulary).
- **wet_strength_treated.why_experts_care**: cut the Trokhan 1982 patent and "micro-regions" detail (kept in the Bounty source entry and in vocabulary as `embossing`).
- **napkin_size.why_experts_care**: cut the exact folded-inch figures beyond the three named categories and the quarter-fold arithmetic (kept in the Napkin sizes source entry and in the facet's `range.typical`).
- **judgements**: the FSC judgement's claim cuts the specific named critics (Greenpeace, FSC-Watch, Simon Counsell) to "named former backers and critics" (kept in the FSC source entry above).

## Single-source numbers

Flagging per the brief's instruction to call these out explicitly:
- Bamboo pulp fibre properties ("closer to softwood pulps") and the three-year harvest age: one source (Wikipedia Bamboo). No second source found; see "Could not reach."
- The napkin quarter-fold size ladder (beverage 5"/luncheon 6.5"/dinner 8.5" folded): three product pages, but one manufacturer family (Hoffmaster/napkins.com). No independent second source (trade body or a second manufacturer stating the convention) found.
- P&G's "consumers primarily preferred absorbency" market-research claim: one source (Wikipedia Bounty, itself reporting P&G's own account). Treated as a judgement about a company's framing of its own research, not a physical fact, and labelled as such.

Everything else that is a number or threshold in `category.json` (the GSM bands, ply ranges, TAD energy cost, lotion dosing, recycled-fibre shortening) has at least two independently fetched sources: either two separate Wikipedia articles (facial_tissue + tissue_paper + Yankee_dryer all independently state overlapping GSM figures) or a Wikipedia article plus the madehow manufacturing primer.

## Script notes

`seed.py`, `validate.py`, `render.py` and `viewer.py` all ran clean with no modification needed; nothing to report as a script problem.

## Corrections (2026-10-26)

- Added `wikipedia_paper_towel` to `references[]` (curl-verified verbatim "Resin size is used to improve the wet strength") and switched `wet_strength_treated.evidence` to cite it instead of `wikipedia_tissue_paper`, which never contained that sentence.
- Added facets `towel_market_class` (domestic/institutional) and `towel_format` (roll/stack), both sourced to the same `wikipedia_paper_towel` sentence, closing the two coverage gaps the blind check found.
- Added facet `tissue_pack_format` (box_dispenser/pocket_pack), sourced to `wikipedia_facial_tissue`'s general "folded and put in pocket-size packages or a box dispenser" line (not the historical Kimberly-Clark pop-up note), closing the facial-tissue pack-format coverage gap.
- Wired `scent_dye_free` into the `paper_napkin` and `paper_towel` branches' `then_ask` (it was previously only wired to `facial_tissue`, contradicting its own `why_experts_care`), and wired the three new facets into all three `product_type` branches' `then_ask`/`makes_irrelevant`.
- Softened `ply.why_experts_care`'s facial-tissue claim from "tops out around 3 ply" to "usually runs 2 to 3 ply", matching `wikipedia_facial_tissue`'s general statement ("these paper types consist usually of 2–3 plies") instead of the historical, single-brand pop-up/3-ply note the blind check flagged.
- Added "Nonwoven" and "Institutional" to `vocabulary`; "Nonwoven" was used undefined in `napkin_material_format`'s definition and the "Airlaid" entry, and "Institutional" is a new term from `towel_market_class`.
- Rejected: the blind check's alternative of a scope note excluding institutional towels instead of a facet — the fix brief requires coverage gaps to become full facets, and institutional/domestic is itself retail-filterable rather than out-of-scope, so a facet is the better fit.