Fujiya Camera Japan Used Gear — Condition-Rank Price Stats in One Call

How much more is an AB body worth than a B body? That spread is what the Japanese used-camera export trade prices off, and it is the only number a plain listing scraper cannot give you. One call per keyword returns it.

This Actor reads Fujiya Camera (フジヤカメラ, Nakano — one of Japan's reference used-camera dealers, whose grading is quoted across the trade) and returns one compact market record per keyword:

• rankStats — the exact population count and median for every condition rank (未使用 / A / AB+ / AB / AB- / B / 現状品 / コレクション). Not a sample: each rank is a filter on the search, so its count is the population count and its median is read out of the population
• rankPremium — how much each rank costs over the cheapest ranked bucket, in percent
• usedPriceJpy — min / Q1 / median / Q3 / max of the whole used result set, with the basis labelled
• newMedianJpy / usedVsNewDiscountPct — the new-price reference the used gear is discounted against
• brandTop, rankMixSampled, plus USD conversion at the current rate
• Optionally every sampled listing (title, brand, price, condition rank, URL)

No login, no API key, nothing stored between runs.

Pricing — from $0.02 per keyword, no subscription

Event	Price	When
Keyword market summary	$0.02	Per keyword analyzed
Individual listing	$0.002	Only if you enable Include individual listings

A default run (1 keyword, summary only) costs $0.02. A keyword that returns zero listings is never charged.

Input

Field	Example	Notes
keywords	["Nikon Z6"]	Maker, model or mount. Each keyword costs $0.02
conditionRanks	["1001","1003"]	Optional: 1000 未使用/未開封品, 1001 A, 1002 AB+, 1003 AB, 1004 AB-, 1005 B, 1006 現状品, 1007 コレクション. Empty = the whole ladder
includeNewReference	true	Also measure the new/outlet price so the used discount can be reported
maxItemsPerKeyword	50	Listings kept for the brand mix (30–300, 50 per page). Does not affect the rank counts, rank medians or quartiles
includeIndividualItems	false	Enable to also get each sampled listing as a record
convertToUsd	true	Adds USD stats at the current exchange rate

Why the rank numbers are exact, and where the basis is labelled

The condition rank is a search filter, not a label the Actor has to infer. So for each rank:

Situation	How the median is obtained	medianBasis
The rank has ≤50 listings	The page is the population — its median is the population median	exact
The rank has more	One extra request jumps to the page holding rank ceil(count/2) of a price-ascending sort — that listing is the population median	population_quantiles
Time budget spent, or the page could not be read	Median of the listings that were inspected	sample

The same three-way labelling applies to usedPriceJpyBasis for the overall used quartiles. The fallback is the point: if the shop ever changes its sort behaviour, the Actor stops claiming population figures instead of quietly returning wrong percentiles.

brandTop and rankMixSampled are always sample-based (sampledListings tells you the size).

Output example (type: "market_summary")

{
  "type": "market_summary",
  "keyword": "Nikon Z6",
  "totalListingsFound": 1325,
  "usedListingsFound": 665,
  "newListingsFound": 660,
  "sampledListings": 50,
  "usedPriceJpy": { "min": 3300, "q1": 39800, "median": 104500, "q3": 199100, "max": 748000, "count": 665 },
  "usedPriceJpyBasis": "population_quantiles",
  "rankStats": [
    { "rank": "unused", "rankLabel": "未使用/未開封品", "count": 12,  "medianJpy": 268400, "medianBasis": "exact" },
    { "rank": "A",      "rankLabel": "A",              "count": 88,  "medianJpy": 179300, "medianBasis": "population_quantiles" },
    { "rank": "AB",     "rankLabel": "AB",             "count": 234, "medianJpy": 119900, "medianBasis": "population_quantiles" },
    { "rank": "B",      "rankLabel": "B",              "count": 41,  "medianJpy": 74800,  "medianBasis": "exact" }
  ],
  "rankPremium": { "baseRank": "B", "baseMedianJpy": 74800, "premiumPct": { "AB": 60.3, "A": 139.7, "unused": 258.8 } },
  "newMedianJpy": 289080,
  "usedVsNewDiscountPct": 63.9,
  "brandTop": [["Nikon", 47]],
  "checkedAt": "2026-08-01T05:31:12.004Z",
  "sourceUrl": "https://www.fujiya-camera.co.jp/shop/goods/search.aspx?search=x&keyword=Nikon+Z6&ps=50&p=1&sort=4&category=20"
}

A caveat worth stating plainly

usedVsNewDiscountPct can come out negative. Measured for Nikon Z6 on 2026-08-01: used median ¥280,500 against a new median of ¥222,750, i.e. −25.9%. That is not a bug — a keyword that names a body also matches accessories (filters, straps, flashes) whose new prices are low, so the "new" side of a loose keyword is not the same product as the "used" side.

Use a keyword specific enough to name one product when you want the discount to mean something, and read newListingsFound (7 in that measurement) as the sanity check. The Actor reports the number rather than hiding it.

What this Actor does not do

• No listing dumps by default. The product is the statistic; individual listings are opt-in and separately priced
• No login-only data. Everything comes from public search pages
• No stored dataset. Every run fetches live; nothing is retained between runs
• No PII. The shop is a single dealer; there are no individual sellers to expose

Notes on the data

• sort is always sent explicitly. The shop's default order is おすすめ順 (merchandising-weighted) and would skew every price statistic
• Ranks are the shop's own ladder, read from its secondhand_state filter. A listing occasionally carries a rank that is not on the filter ladder; conditionRank then reports the shop's wording verbatim rather than forcing it into a bucket
• Prices are asking prices including tax as displayed by the shop
• A run costs roughly 1 + 4 + (1–2 per rank) + 1–2 requests per keyword, throttled at 1.2 s. Multi-keyword runs carry a wall-clock budget and set truncatedForTimeLimit rather than silently truncating