# Does Lebanon pay more for the same Zara item? Method Measured 8-9 October 2026 (UTC). Everything below is reproducible with `fetch.py` in this folder (stages: `sample`, `fetch`, `fx`, `analyze`, `handcheck`, `repass tr`). Numbers quoted here come from `summary.json`. ## What was compared The same Zara product, in the same colour, on Zara's own public websites in Lebanon (lb/en), the United Arab Emirates (ae/en), Spain (es/en), Turkey (tr/en) and the United States (us/en), at its regular (non-sale) shelf price, converted to US dollars. ## Sample - Source: closet.ai's daily scrape of Zara Lebanon (`closet/data/products/zara.json`, 27,888 records). - Eligible: records with `listed == true`, `last_seen == 2026-10-08` and `was == null` (not on sale in Lebanon): 15,102 colour records, 11,202 distinct products. - One colour was drawn at random per product, so no product counts twice. - Stratified by gender (Women, Men, Kids) x category (60 strata), with proportional allocation (largest remainder, at least 1 per stratum), then a simple random draw inside each stratum. Seed 20261009. 450 products: 201 Women, 105 Men, 144 Kids. List: `sample.csv`. - Results are unweighted medians over the sample. Because allocation is proportional, the sample mirrors the Lebanon catalogue's mix of genders and categories. ## Matching rule (like for like) - Product key: the 8-digit code in the product URL (`p06095982`). It is a 1-digit prefix (0, 1 or 2) + 4-digit model + 3-digit quality. closet.ai's `ref` field ("6095/982") drops the prefix, so it is not unique on its own. - Colour key: the 3-digit colour code, the suffix of the closet record id (`zara-545400334-250` -> 250). The same digits follow the product code in Zara's image names (`06095982` `250`) and in Zara's colour reference (`C06095982` `250` `015-I2026`). - A country row is a match only when that country's page returns the same product code and a colour with the same 3-digit id whose reference starts with `C`. Otherwise it is recorded as a failure with the reason. - Zara's internal product ids were identical across all five stores in every read checked, which supports this being the same item. ## Fetching - URL: `https://www.zara.com/{cc}/en/{slug}-p{code}.html?ajax=true`. Zara returns the product page's data as JSON. Plain HTTPS (Python httpx), a desktop Chrome User-Agent and `Accept-Language: en-US`. No login, no browser, no API keys. - One request at a time, 1.5-2.5 s random pause, back-off on 403/429/5xx (none happened: 0 refusals in 3,055 HTTP requests for 2,740 page reads, redirects included). Wrong or empty slugs are answered with Zara's ajax redirect (HTTP 278, `{"location": ...}`), which is followed. A product not sold in a country answers HTTP 410. - Each item was fetched in all five countries back to back (Lebanon first), so the Lebanon price is from the same minutes as the foreign prices. Main pass: 2026-10-08 23:00 UTC to 2026-10-09 00:40 UTC. - Price: the colour's `price`, which Zara sends as the amount x 100 in every currency (Lebanon 490000000 = 4,900,000 LBP). Currency from the page's analytics block. Per-size prices were also read: none of the matched rows had different prices across sizes. - Sale: a row is "on sale" when the colour or any of its sizes carries an `oldPrice`. Sale rows are kept in `prices.csv` but left out of the regular-price comparison. On sale: 2 rows (US), 0 elsewhere. - Lebanon's same-day page price equalled closet.ai's recorded price for all 448 matched Lebanon rows. ## Currency Zara Lebanon's website prices in Lebanese pounds (LBP), not dollars. Its prices are converted at Banque du Liban's official rate. | Currency | USD per unit | Source | |---|---|---| | LBP | 1 / 89,500 | Banque du Liban, Exchange Rates, US Dollar series (USD/LBP), 89,500 every month Aug 2024 to Jul 2026 (latest published), read 2026-10-08 | | AED | 1 / 3.6725 | UAE dirham peg | | EUR | 1.1186 | ECB euro reference rate, 2026-10-08 | | TRY | 1.1186 / 55.0523 = 0.020319 | ECB euro reference rates (USD and TRY), 2026-10-08 | - ECB: the latest reference date on or before the first fetch (published 2026-10-08, ~16:00 CET); the 9 October rates were not out during the run. - Market check for LBP: Wise mid-market 89,550 per USD (read 2026-10-08 23:05 UTC), 0.06% from the official rate. Zara's own internal rate in the pages implies 89,577 LBP per USD. Neither changes any result. - All rates, URLs and read times: `fx.json`. ## Taxes - Prices are shelf prices as shown. US prices exclude sales tax (added at checkout, varies by state). Spain (21% VAT), the UAE (5% VAT) and Turkey include VAT in the shelf price. - Lebanon has 11% VAT. The Zara Lebanon site does not say whether its shelf prices include it; we compare the shelf price as shown and do not claim either way. - Secondary comparison: Spain and the UAE with their VAT taken out (price / 1.21 and / 1.05), Lebanon left as shown. This is a "before tax abroad" view, not a like-for-like tax comparison. ## Results (regular prices, Lebanon / other country, in USD) "More expensive" means the Lebanon price is more than 1% above; "equal" means within 1%. | vs | n | Median LB premium | Mean | LB dearer | LB cheaper | Equal | |---|---|---|---|---|---|---| | UAE | 443 | +0.3% | +1.1% | 49% | 35% | 16% | | Spain | 438 | +63.2% | +63.4% | 100% | 0% | 0% | | Turkey (prices after 9 Oct repricing) | 435 | +25.3% | +27.2% | 100% | 0% | 0% | | Turkey (prices before it, main pass) | 435 | +34.0% | +33.3% | 100% | 0% | 0% | | United States | 439 | -5.7% | -3.5% | 21% | 67% | 13% | - Ex-VAT: vs Spain ex-VAT, median +97.5%; vs UAE ex-VAT, median +5.3% (LB dearer for 80%). - By gender, US: Kids +3.6% (LB dearer for 54%), Men -7.1%, Women -8.6%. UAE: Kids +4.2%, Men -2.8%, Women -0.1%. Spain and Turkey: every gender and category is dearer in Lebanon. Full breakdowns by gender and category: `summary.json` (`by_gender`, `by_category`). - 426 items matched in all four comparisons; their medians are the same to within 0.2 points (`matched_in_all_countries`). - Top 15 Lebanon premiums per country, with both prices and URLs: `summary.json` (`top15_lb_premium`; Turkey's list uses main-pass prices). ## Hand check and the Turkey repricing - 20 random matched items were re-read later in a separate pass (00:41-00:42 UTC on 9 October), Lebanon plus one other country each (`handcheck.csv`). Lebanon 20/20, UAE 5/5, Spain 3/3 and US 4/4 identical. Turkey: 5 of 8 had changed, all upward (4-10%). - Investigation: Turkey was re-read in full (`repass tr`, 00:43-01:00 UTC). 279 of 437 matched Turkish prices had gone up between the passes (median +7.5% among those that changed), none down. Items first read before about 00:15 UTC changed in most cases; items first read after 00:20 UTC had not changed (1 of 93). So Zara Turkey raised prices at about 00:15-00:20 UTC on 9 October (03:15-03:20 Turkey time), in the middle of the main pass. - The repass prices are the current ones, so the Turkey headline uses them (+25.3%). The pre-repricing number (+34.0%) is kept for transparency. The TRY rate is the ECB's from 8 October, before the repricing. ## Failures (out of 450 per country) | Country | Matched | Not sold there (410) | Colour not offered | On sale | LB side not comparable | |---|---|---|---|---|---| | Lebanon | 448 | 0 | 2 | 0 | - | | UAE | 444 | 4 | 2 | 0 | 1 | | Spain | 439 | 10 | 1 | 0 | 1 | | Turkey | 437 | 13 | 0 | 0 | 2 | | United States | 443 | 7 | 0 | 2 | 2 | "LB side not comparable" = matched abroad, but the item's colour was no longer offered in Lebanon on 9 October. ## Caveats - One snapshot (one night). Prices move: Turkey repriced during the run. - Shelf prices only: no shipping, duties, delivery fees, or US sales tax. - The sample is of Lebanon's live full-price catalogue. Items Zara sells abroad but not in Lebanon are not in it. - Lebanon's dollar figure depends on the LBP rate; the official rate and the market rate were 0.06% apart on the day. - Median over items, not weighted by what people buy. ## Files - `fetch.py`: the script (config at the top). - `sample.csv`: the 450 sampled items. - `raw.jsonl.gz`, `handcheck_raw.jsonl.gz`, `repass_tr.jsonl.gz`: each read as received (trimmed to the product's colours, prices, sizes and analytics). - `prices.csv`: one row per item x country per pass (`pass` = main or repass), including failures, with URL, HTTP status and UTC timestamp. - `fx.json`, `summary.json`, `handcheck.csv`.