# Phase 0 — Environment, model and source audit (WP38)

*Written 2026-11-01. All figures below were measured on the authorised snapshot or on the production node with read-only queries; nothing was modified in production.*

## 0.1 Environment

| Item | Value |
|---|---|
| Working directory | `~/Desktop/UQO/Recherche/Recherche-UQO/economie/economic-geography-digital-production` (git, local `main`; remote to be created on spbgit as `wp38_uqo.git`) |
| Laptop | Apple Silicon, 18 cores, 48 GB RAM, 1.0 TB free; macOS 27 |
| Cluster | MacLustr (29 Macs, 401 cores) reachable over SSH; only M4M64a (Fabri-Ka host) was used in Phase 0 |
| Python | `uv` venv `.venv`, Python 3.12, pinned via `pyproject.toml` (pandas, pyfixest, statsmodels, networkx, geopandas, duckdb, rapidfuzz, …) |
| LaTeX | TeX Live (`latexmk`, `pdflatex`), `pandoc` |
| Network | Outbound HTTPS allowed. Crossref, Statistics Canada WDS, ISQ and Données Québec are reachable |
| API keys | `.env` (git-ignored) holds one OpenAI key and one Anthropic key inherited from WP37; never printed |

### 0.1.1 Model inventory (official provider endpoints, tiny test calls)

See `reports/model_inventory.md` / `logs/model_inventory.json` (checked 2026-11-01T17:42Z). OpenAI `GET /v1/models` lists 140 IDs; `gpt-5.5`, `gpt-5` and `gpt-5-mini` answered 200 (8-token budget fully consumed by reasoning, so empty text — expected for reasoning models with a tiny budget); `gpt-5.1` returned 400 ("could not finish the message" under the 8-token budget). Anthropic lists 14 IDs; `claude-fable-5-1`, `claude-opus-5-5`, `claude-sonnet-5-5`, `claude-haiku-5-5` all answered "OK" in about 1 s. Total cost of the inventory: below 0.002 USD. LLMs will be used only for (a) second annotation of a product-matching validation sample and (b) optional category-reclassification audit; every call will be logged in `logs/run_ledger.jsonl` with its cost. No empirical result depends on an LLM.

## 0.2 Source identification: Fabri-Ka

Fabri-Ka (www.fabri-ka.com) is an aggregator built and operated by the author (Groupe Ka). It **does not sell**: it discovers Quebec online stores from certification registries, sector associations, regional agri-food brands and marketplaces (`deliverables/research_report.md`), verifies each domain, classifies the *store* by origin class A–E with retained textual evidence, and then syncs each store's **public catalogue endpoint** (Shopify `/products.json`, WooCommerce Store API, Wix GraphQL, Squarespace `?format=json`, Square, Votresite, generic JSON-LD) roughly every 6 hours (PM2 process `fabri-ka-sync`). Compliance per source is documented in `docs/CONFORMITE.md` (reviewed 2026-08-19/21): public endpoints only, nothing behind authentication, polite rates, removal on request.

### 0.2.1 Access and authorisation

- **Authorised access:** the author owns the platform and the database. The research copy is a first-party export; no scraping of fabri-ka.com and no use of its `/api/` (which `robots.txt` disallows for crawlers) beyond **one** control request to `/api/stats` on 2026-11-01T17:37Z to date-stamp the public counters (products 390 108, stores_live 1 569, stores_registry 4 082, regions 17). The mandate's "approximately 390 456 products, 1 572 stores" was therefore an interface figure from a slightly earlier moment; the counts drift by a few hundred per 6-hour cycle.
- **Snapshot:** `sqlite3.Connection.backup()` of `M4M64a:~/fabri-ka/data/fabrika.db` on **2026-11-01T17:38:30Z**, application commit `689c7bb1…`; tables `users` (2 rows) and `inscriptions` (1 row) — the only personal data — and the derived FTS index were dropped on the node before transfer. SHA-256 `8d963ba9…98fb`, 1.24 GB, stored in `data/raw/` (git-ignored). The discovery registry `stores.json` and `collections_taxonomy.json` were copied with checksums (`data/raw/README.md`).
- **Production untouched:** all queries used `mode=ro`; the WAL file was not modified.

### 0.2.2 Schema (three tables kept)

| table | rows | key fields |
|---|---|---|
| `stores` | 4 082 (1 569 enabled with ≥ 1 published product; 132 enabled with 0; 2 381 disabled = directory-only or dead) | `id` (canonical domain), `name`, `platform`, `city`, `region` (17 canonical names or blank), `origin_class` A/C/E, `origin_confidence`, `origin_evidence` (text), `categories` (JSON), `discovery_sources` (JSON), `language`, `postal_prefix` (RTA/FSA), `lat`,`lng` (FSA centroid via zippopotam/GeoNames, precision "fsa"), `store_kind` (revendeur/collectif/blank), `certifications` (JSON: Aliments du Québec, Les Produits du Québec, CARTV bio…), `product_count`, `fail_streak`, `presumed_closed`, `shipping_info` |
| `products` | 465 388 (390 108 `active=1 ∧ listing_status='published'`) | `uid` (20-hex hash of store+external id), `store_id`, `external_id`, `url`, `title`, `description`, `price` (min variant), `price_max`, `compare_at_price`, `currency`, `images` (JSON), `category` (20 canonical keys, inferred by keyword rules `fabrika/schema.py::infer_category` from product type, tags, title and option names), `product_type` (merchant's own), `tags` (JSON), `vendor` (merchant's brand field), `available`, `content_hash`, `first_seen`, `last_seen`, `active`, `miss_count`, `details` (JSON: variants with price/sku/grams/available, options, attributes, categories), `listing_status` (published / excluded:{carte_cadeau, atelier_cours, abonnement, billet, test} / quarantine:{prix_hors_bornes, titre_vide}), `price_on_request` |
| `sync_log` | 577 773 (ok 572 749, error 4 972, presumed_closed 52) | one row per store per sync: `ts`, `store_id`, `found`, `added`, `updated`, `removed`, `status`, `message` |

**There is no price-history table.** The database upserts by `content_hash` and keeps only the current price per listing; `sync_log` records counts, not prices. Consequently every price analysis in WP38 is **cross-sectional at the snapshot date**; temporal analysis is limited to listing entry/exit (`first_seen`/`last_seen`) and store-level sync counts.

### 0.2.3 Temporal coverage

- `first_seen` runs 2026-08-12 → 2026-11-01 (81 days). 291 125 listings carry `first_seen` = 2026-08-12 (initial load) and a further 48 820 + 30 173 on 2026-08-19/23 (expansion waves); 1 266 stores were onboarded in ISO week 32 and 328 in week 33; only 15 stores were first observed later. Hence **store onboarding ended by 2026-08-24**, and product-level entry after that date reflects merchant catalogue changes (new listings, connector fixes) rather than platform expansion — this is the only usable "dynamic" margin and it must be audited connector by connector before any H4 statement.
- Sync cadence: median 6.3 h between successive syncs of a store (p10 1.2 h, p90 8.8 h; 200-store sample, last 30 days). A listing becomes `active=0` after 2 consecutive misses (`MISS_GRACE = 2`, ≈ 12 h). 73 618 published listings are inactive at the snapshot (last seen Aug 6 970, Sep 31 642, Oct 35 222). A disappearance is **not** a sale (sold-out products usually stay listed with `available=0`: 48 997 active listings).

### 0.2.4 Field-level quality and missingness (active published listings, N = 390 108)

| issue | measurement |
|---|---|
| Price missing | 14 786 (3.8 %) `price IS NULL` (all flagged `price_on_request`); 31 quarantined `prix_hors_bornes` (> 100 000 CAD) excluded upstream |
| Currency | CAD 389 378; USD 578 (8 stores); EUR 112; GBP 40 — non-CAD stores to be excluded from price analyses (12 stores) |
| Price distribution | p1 1.32, p25 12.95, median 25.00, p75 69.00, p95 766, p99 3 745 CAD; max 95 999.99 |
| Variants | `price_max > price` for 57 728 listings (14.8 %) — the listing price is the **minimum variant price**; `details.variants` carries per-variant prices for Shopify/WooCommerce |
| Compare-at (strike-through) price | 33 311 listings (8.5 %) have `compare_at_price > price` (on-sale indicator, merchant-declared) |
| Availability | available = 1: 336 038; = 0: 48 997; NULL: 5 073 |
| Category | 20 keys; `autre` = 46 094 (11.8 %); rule-based inference, no manual validation yet → WP38 will audit a stratified sample and report category precision |
| Brand / vendor | `vendor` non-empty for 290 827 (74.5 %); it is the merchant's own brand field (often the store name) |
| Product type / tags / description / images | non-empty for 305 860 / 261 421 / 339 265 / 360 508 |
| SKU / GTIN | SKU present inside `details` for Shopify/Woo variants; **no GTIN/UPC field** → cross-merchant matching must rely on vendor + normalised title (+ SKU where present) |
| Within-store duplicate titles | 7 411 (store, title) pairs with > 1 listing (variants listed as separate products, colour/format) |
| Cross-store identical titles (len > 15) | 2 595 distinct titles covering 5 961 listings — first candidate pool for H5 |
| `external_id` collisions across stores | 15 033 — platform-local integer IDs, **not** product identity (ignored) |

### 0.2.5 Merchant-level quality

| issue | measurement |
|---|---|
| Region | **346 of 1 569 live stores (22 %) have a blank region, carrying 159 096 listings (41 %)**, including the four largest catalogues (qublivre.ca 24 946 books; ici-la.co 21 412 souvenirs; remorquetrailer.com 19 482; beteaunaturel.com 9 910). Of these 346, 117 have an FSA centroid (`lat/lng`) → region recoverable by point-in-polygon against the official administrative-region layer; 229 have no geographic field at all → to be resolved from `stores.json` (`qc_postal`, `city` mentions) or left as "unknown region" and excluded from regional analyses (sensitivity: included as an 18th pseudo-region) |
| Region provenance | assigned by (1) regional directory source (certain), (2) regions mentioned on the site, (3) FSA → region map; coordinates = FSA centroid (neighbourhood precision), **never the civic address** |
| Origin class | A 1 526 stores (322 007 listings), C 12 (63 885), E 31 (4 216); B and D unused. Class is a **store-level prior from the discovery source** (e.g. PPAQ maple producer = A 0.95; Les Produits du Québec retailer = C 0.75) adjusted by site signals; `origin_evidence` text retained. Many A entries read "aucun signal de localisation québécoise détecté sur le site (à vérifier)". It is **not** product-level manufacturing provenance |
| Certifications | 718 live stores carry at least one structured certification (Aliments du Québec, Les Produits du Québec, CARTV organic, …) — third-party **store/enterprise** certification, usable as a store-level provenance indicator with its own caveats |
| Platform | shopify 696 (273 504 listings), woocommerce 498 (91 100), wix 244 (8 798), votresite 42, square 36, squarespace 23, generic 20, lightspeed 5, others ≤ 2 |
| Catalogue size | median 36 listings per store, p90 324, p99 3 997; top-10 stores hold 139 735 listings (36 %) |
| Store kind | revendeur 23, collectif 3 (marketplaces such as epipresto.ca), hors_mission 1, blank 1 542 |
| Sync health | 26 stores with an error status in the last 7 days; `stores.product_count` equals the live published count for all 1 569 stores (consistency check passed) |

### 0.2.6 Known selection mechanisms (from the platform's own documentation)

Discovery favours certified, association-member and search-indexed businesses; under-represents CMAQ artisans, industrial manufacturers selling direct, Instagram/Facebook-only micro-shops, Nord-du-Québec, Etsy sellers (excluded by Etsy ToS), ASC farms; Wix stores were only connected in a later wave; Ecwid storefronts cannot be connected; Montréal was deliberately topped-up via a dedicated source (`acheterquebecois_mtl`, `modemtl`). Connector caps (`max_pages` 200) truncate the very largest catalogues. These mechanisms are region- and category-correlated and are central to the representativeness audit (Phase 5).

## 0.3 Feasibility verdict by hypothesis

| hypothesis | feasible with this source? | conditions |
|---|---|---|
| H1 regional specialization (RAA) | **Yes** | after region recovery; merchant-weighted and largest-merchant-excluded variants mandatory |
| H2 diversity vs regional scale | **Yes** (n = 17 regions → low power; report as descriptive association with bootstrap CIs) | external population/business counts from ISQ/StatCan |
| H3 concentration / long tail | **Yes** | rank-size of merchant catalogues, listing shares |
| H4 related diversification (dynamic) | **Partially / exploratory only** | no store onboarding after week 33; only within-store listing entry over ~10 weeks; no cross-snapshot category entry at region level beyond a handful of events → mark as exploratory, likely inconclusive |
| H5 price dispersion of matched products | **Yes, cross-sectional only** | no price history; matched products via vendor + title + variant; merchant/region collinearity: one merchant = one region → region effect identified only across merchants |
| H6 local-origin premium | **Not testable at product level** (no product provenance) | store-level certification contrast reported as a descriptive, clearly-labelled proxy, not as H6 |
| H7 digital-assortment complexity | **Yes** (17 × 20 matrix; small) | null models and merchant-preserving permutations required; eigenvector interpretation guarded |

## 0.4 Blockers and decisions carried forward

1. No price history → temporal price analysis blocked (recorded in `reports/limitations.md`). A prospective collection plan is written but not executed.
2. Region unknown for 22 % of live stores → Phase 2 region-recovery procedure with quality flags; all regional results reported with and without recovered stores.
3. Origin class is store-level → wording throughout: "online assortment of Quebec-based merchants", never "Quebec production".
4. Category labels are rule-based → stratified audit (manual + LLM second annotation, logged cost) to estimate confusion rates and propagate uncertainty.
5. Public counters vs cleaned sample: 390 108 published active listings at the snapshot; the research sample will be smaller after exclusions (non-CAD, missing price, unknown region, duplicates) and the full attrition is reported in `reports/sample_flow.md`.