Isetan Mitsukoshi Online Store Scraper

Scrape products from the Isetan Mitsukoshi department store online shop at https://www.mistore.jp/shopping and get a clean, flat record per product: identity, brand, department and breadcrumb trail, price in Japanese yen, sale flags and the sale window, stock, colour options, the full image gallery, size and colour variants with per variant price, barcode and stock count, the specification list, shipping and gift terms, payment methods, allergen panel, and customer reviews where the store has them.

This is a genuine department store catalogue: womenswear, menswear, baby and kids, bags, shoes, home, kitchen and art, food and gifts are all covered by the same actor.

---

What you get

49 fields per product. Everything below is a real field on every row, so a dataset never changes shape between a light run and a full one.

Group	Fields
Identity	productId, title, brand, brandCode, brandUrl, sku, productNumber, url
Category	topCategory, categoryPath, categoryCode, breadcrumbs
Price	price, priceMin, priceMax, originalPrice, discountPercent, currency
Sale and stock	onSale, saleStartsAt, saleEndsAt, inStock, availability, stockNote, badges, badgeCodes, cardholderOnly, digitalProduct
Media	image, images, thumbnails, videos
Variants	variants[] with color, colorCode, size, sizeCode, barcode, price, originalPrice, discountPercent, stock, inStock, onSale, image, saleStartsAt, saleEndsAt
Detail	description, specifications[], allergens[], notices[]
Delivery and terms	shippingFee, shippingNote, deliveryNotes, giftOptions[], paymentMethods
Reviews	rating, reviewCount, ratingBreakdown, reviews[] with author, rating, title, text, postedAt, postedTime, helpfulCount, totalVotes, images
Run and change tracking	searchMode, scrapedAt, changeType, changedFields, firstSeenAt, lastSeenAt

---

Modes

Search

Give one or more keywords, one or more departments, or both. Each is covered separately, across all of its results.

{
  "mode": "search",
  "queries": ["バッグ"],
  "categories": ["womens"],
  "sortBy": "price_asc",
  "badges": ["sale"],
  "minPrice": 5000,
  "maxPrice": 50000,
  "fetchDetails": true,
  "maxItems": 100
}

Departments accept a friendly name (womens, mens, baby_kids, bags, shoes, home_kitchen_art, food, events) or a store category code such as 28_011500 for a narrower sub-department.

The store files bags and shoes under each audience rather than as one store wide department, so bags browses the women's, men's and kids' bag departments and shoes browses the women's, men's, kids' and baby shoe departments. Each is covered in full, and maxItems is shared evenly between them.

If you paste your own category code, check it against the department you meant. This store does not reject a code it does not recognise — it quietly returns the entire catalogue instead, so an unrecognised code looks like a working department that happens to be very large.

URL

Paste links and let the actor route them. A listing link is covered in full; a product link or a bare product id returns that one product with full detail, and is always returned regardless of the filters.

{
  "mode": "url",
  "urls": [
    "https://www.mistore.jp/shopping/productlist?cgid=28_011500",
    "https://www.mistore.jp/shopping/search?q=%E3%83%AF%E3%82%A4%E3%83%B3",
    "https://www.mistore.jp/shopping/brand/list?brand=002383",
    "https://www.mistore.jp/shopping/product/900000000000000003062447.html",
    "900000000000000003626529"
  ],
  "fetchDetails": true
}

---

Filters

Every filter is the store's own refinement, sent with the request rather than applied afterwards, so it narrows the crawl itself and you are not billed for products you filtered out. They apply to a keyword search, a department and a pasted listing link alike.

Filter	Values
sortBy	recommended, newest, price_asc, price_desc
badges	sale, new_arrival, point_bonus, limited_time, store_exclusive, online_exclusive, message_card
colors	white, black, beige, green, blue, purple, grey, brown, yellow, pink, orange, red, silver, gold, other
brandCodes	store brand codes, e.g. 002383
minPrice / maxPrice	yen
includeOutOfStock	off by default, matching the store

---

Output sample

{
  "productId": "900000000000000003062447",
  "title": "型押しハンドルバッグ(9520BK17)",
  "brand": "MONO COMME CA (Women)/モノコムサ",
  "brandCode": "022826",
  "url": "https://www.mistore.jp/shopping/product/900000000000000003062447.html",
  "topCategory": "レディース",
  "categoryPath": ["レディース", "バッグ", "ハンドバッグ"],
  "categoryCode": "28_011504",
  "price": 4829,
  "priceMin": 4829,
  "priceMax": 4829,
  "originalPrice": null,
  "discountPercent": null,
  "currency": "JPY",
  "onSale": true,
  "inStock": true,
  "availability": "InStock",
  "variants": [
    {
      "variantId": "100000000000000375697687",
      "color": "ブラック",
      "colorCode": "10",
      "size": "ｻｲｽﾞ ﾅｼ",
      "barcode": "4549244554673",
      "price": 4829,
      "originalPrice": null,
      "stock": 2,
      "inStock": true,
      "onSale": true
    }
  ],
  "specifications": [
    { "name": "素材", "value": "ブラック/グレー/ダークグレー：本体:合成皮革裏地:ポリエステル付属:合成皮革" },
    { "name": "原産国", "value": "中国" }
  ],
  "shippingFee": 715,
  "giftOptions": [{ "name": "ギフト包装", "available": true, "note": "対応可" }],
  "paymentMethods": ["エムアイカード", "他のクレジットカード", "d払い"],
  "badges": ["SALE", "配送日指定不可"],
  "rating": null,
  "reviewCount": 0,
  "reviews": [],
  "scrapedAt": "2026-08-27T07:33:55Z"
}

---

Reviews

Customer reviews are available and are extracted when you turn on fetchReviews. A reviewed product returns its aggregate rating, its review count, a per star breakdown, and a row per review with the author handle, the post date and time, the star rating, the title, the full text and the helpful vote tally.

{
  "rating": 5.0,
  "reviewCount": 1,
  "ratingBreakdown": { "0": 0, "1": 0, "2": 0, "3": 0, "4": 0, "5": 1 },
  "reviews": [
    {
      "author": "<reviewer handle>",
      "rating": 5.0,
      "title": "可愛くてお洒落",
      "text": "娘へのプレゼントに。ジャカード織りでしっかりしていて…",
      "postedAt": "2026-08-19",
      "postedTime": "12:28:42",
      "helpfulCount": 1,
      "totalVotes": 2,
      "images": []
    }
  ]
}

Every review a product declares is returned, not just the first page. The store's review surface serves three reviews per page and states its own position in a footer ("showing 4 to 6 of 13"). The actor walks forward on that footer until it holds the full count, so reviewCount, ratingBreakdown and the length of reviews always agree with each other. ratingBreakdown tallies exactly the reviews in the row: if you set maxReviewsPerProduct lower than a product's total, the breakdown counts what you got, while reviewCount still reports the store's own total. The buckets run 0 to 5 because the store really does accept a zero star review, and a one to five histogram would drop those from a row that still lists them.

Be aware that reviews on this catalogue are sparse. Checked live across bags, food, cosmetics, homeware and wine, only a small minority of products carry a review at all, and most reviewed products carry exactly one. Products with more do exist, so the walk is real rather than decorative. Most products come back with reviews: [], reviewCount: 0 and rating: null. Those are honest empty values, not fabricated zeros, and no review is invented for a product that has none. A product with one review costs exactly one request and never speculatively probes a second page. fetchReviews is off by default because it costs at least one extra request per product.

Discounts and original prices

originalPrice and discountPercent are extracted at both the product level and the variant level whenever the store publishes them. At the time of writing the store does not publish them. It flags products as SALE and shows the reduced price, but sampled live across 100 sale flagged and limited time products in six departments, the list price and the discount rate were empty on every single one. So onSale is reliable, and originalPrice and discountPercent are honestly null rather than being back-computed from the current price. If the store starts publishing them, for example during a seasonal clearance, they come through automatically.

Countries

This is a Japan only storefront and the actor has no country input. The store publishes no country or region selector and no currency selector; every price is in Japanese yen, there are no alternate locale pages, and the store's own help pages state that it does not ship internationally. The language switch on the site translates the same Japanese catalogue at the same yen prices rather than opening a different country's shop. The Isetan Mitsukoshi group does run department stores in other countries, but those are separate businesses on separate sites, not another storefront of this one. Inventing a country input with a single working value would have been misleading, so there is none.

Stock status, and how department listings are read

inStock is true, false, or null meaning the store did not publish it at that point — never a guess.

The two listing surfaces differ, and it matters:

• A keyword search publishes a stock flag for every product on the page, so search rows carry a real true or false straight from the listing.
• A department listing publishes no stock flag at all. It renders six products into the page and serves the other 54 as placeholder tiles that carry only a product id and an ex tax price — no title, no brand, no image, no stock flag.

The actor resolves those placeholders through the store's own tile endpoint, the same request the site's page makes as you scroll, which returns the real title, brand, image and tax included display price. That costs one small extra request per placeholder and is bounded by maxItems, so it is never paid for beyond what your run can use.

Two deliberate choices follow from this, both chosen over quietly guessing:

1. A missing stock flag stays null, and the default in stock filter never drops a null. Treating "not published" as "out of stock" would silently discard about 90% of every department page. When fetchDetails is on, the product page settles the stock status and the filter is applied to that real value instead.
2. A placeholder that could not be resolved is never returned or billed. If the tile request fails, the actor has a product id and nothing else. A row of nulls is worse than no row, so it is dropped rather than emitted as a dataset item.

The placeholder's own minprice is deliberately not used as the price: it is an ex tax figure and runs 8% below the shelf price on food and 10% below on general goods, which are Japan's two consumption tax bands.

A note on brand on light runs

The store leaves the brand blank on its own listing tiles for many food, wine and gift products, where the brand is usually part of the title instead. Those rows come back with brand: null on a listing only run. Turning on fetchDetails fills the brand from the product page, where it is always present.

---

Recurring runs

Incremental mode

Turn on incrementalMode for scheduled monitoring. The first run returns everything as NEW; later runs return only NEW, UPDATED and REAPPEARED, each row carrying changeType, changedFields, firstSeenAt and lastSeenAt. Unchanged products are suppressed, so you are not billed for rows you already have. Turn on emitUnchanged or emitExpired only if you also want those returned, and billed.

State is kept per combination of mode, keywords, departments, links, filters and enrichment settings, so two different searches never mix. Set stateKey to name a campaign or to deliberately share state.

EXPIRED rows are only produced after a run has genuinely scanned the whole tracked search. A run stopped by maxItems, by maxPages, by Resume, or one that read nothing at all, leaves the previous state untouched instead of declaring products gone.

Resume

Paste a previous run id or dataset id into resumeFromRunId to continue one interrupted crawl without returning or being charged for what it already collected. Runs also survive a platform migration or a Resurrect on their own via an internal checkpoint.

---

Limits and cost

• maxItems caps the run. 0 means unlimited.
• maxPages caps result pages per keyword, department or link. 0 means unlimited, and the run then ends only when the store runs out of results. Each result page carries 60 products.
• fetchDetails and fetchReviews each add one request per product and together add one Detail enrichment charge per enriched product. Turn both off for the cheapest possible run.
• If enrichment starts failing across a run, it is switched off automatically for the rest of that run: products still come back with their listing fields, and the enrichment charge is not applied to them.

Connection

The default is Apify Proxy with a Japanese exit on the metered residential pool — the only pool that offers a Japanese exit on every Apify plan. If your plan includes a cheaper pool, select it in the Connection section to lower the run cost.

Send results into your apps (MCP connectors)

Optionally pipe results into the apps you already use through Model Context Protocol (MCP) connectors. This is an extra delivery step after the scrape, and it never changes the dataset.

1. Authorize a connector once under Apify, Settings, API and Integrations (Notion, Linear, Airtable or Apify).
2. Select it in the mcpConnectors field. If the picker is empty, you have not authorized a connector yet.
3. For Notion, also set notionParentPageUrl to the URL or id of the page the product pages should be created under. Notion gets a rich page per product; other connectors get a best effort write or digest.
4. Use maxNotifyListings (default 50, range 1 to 1000) to cap how many products are written to each connector per run.

Each connector receives a condensed, human readable summary per product: the title, brand, price, sale flag, stock status and product link. The complete record, including variants, specifications, allergens, delivery terms and reviews, always stays in the Apify dataset. The connection is mediated by Apify's MCP proxy, so this actor never sees your third party credentials. Leave mcpConnectors empty to skip.

Parameter	Type	Default	Description
mcpConnectors	array	(empty)	Connectors to pipe results into. Leave empty to skip.
notionParentPageUrl	string	(empty)	Notion connector only. Page under which product pages are created. Ignored by other connectors.
maxNotifyListings	integer	50	Cap on products written to each connector per run. Does not affect the dataset.

---

Local development

python3.12 -m venv .venv
.venv/bin/pip install -r requirements.txt
PATH=.venv/bin:$PATH apify run --purge
.venv/bin/python -m pytest tests/ -q