Auctionet Scraper — lots, hammer prices and auction house contacts

Every lot on auctionet.com as one flat row: what it is, what it was expected to make, every bid it took, what it actually sold for — and the phone number, email, website, street address and coordinates of the auction house selling it.

Measured on the live site on 2026-09-18 from this machine:

Live lots open for bidding	39,189
Sold lots in the hammer-price archive	3,938,923, back to 2014
Auction houses in the public directory	86 across Sweden, Denmark, Finland, Germany, Spain and the UK
Houses with live lots right now	84
Houses with archive lots	91
Currencies	SEK, EUR, GBP, DKK — amounts are never converted
Price	$3.00 per 1,000 lots, no run-start fee

What one row is

One row = one lot. Two corpora, chosen with the Live lots or the sold archive input:

• Live — the 39,189 lots currently taking bids. Changes by the minute. For dealers sourcing stock, and for anyone watching a category for an opportunity.
• Sold — the 3,938,923-lot archive of lots that have already hammered, each with its realized price. Static. For price databases, appraisers, insurers and anyone building comps.

Field fill, counted — not claimed

Counted by the validator that ships in this repo, on two real captured pages of 200 lots each (fixtures/items_live_page1.json and fixtures/items_sold_company9_page1.json, served by auctionet.com on 2026-09-18). "", "-", "N/A" and <p></p> are not counted as filled. A real 0 or false is.

Field	Live (n=200)	Sold (n=200)	What it is
lotId	200/200 100.0%	200/200 100.0%	Auctionet's lot id. Stable, and the dedupe key.
title	200/200 100.0%	200/200 100.0%	Lot title, in the auction house's own language.
longTitle	200/200 100.0%	200/200 100.0%	The site's own "<id>. <title>" form.
lotUrl	200/200 100.0%	200/200 100.0%	Full link to the lot page.
catalogNumber	0/200 0.0%	0/200 0.0%	The house's printed catalogue number, when it publishes one.
auctionId	200/200 100.0%	200/200 100.0%	Id of the auction the lot belongs to.
eventId	0/200 0.0%	0/200 0.0%	Id of the themed event/catalogue sale, when the lot is in one.
lotType	200/200 100.0%	200/200 100.0%	Sale type as the API reports it (online).
lotStatus	200/200 100.0%	200/200 100.0%	live, sold or ended.
state	200/200 100.0%	200/200 100.0%	The source's own state string (published / sold).
isEnded	200/200 100.0%	200/200 100.0%	Whether bidding has closed.
isHammerAuctionLot	200/200 100.0%	200/200 100.0%	Auctionet's hammered flag. It marks an in-room hammer auction lot, not "this lot sold" — see Honest limits.
description	200/200 100.0%	200/200 100.0%	Lot description as served: HTML, native language, untranslated.
descriptionText	200/200 100.0%	200/200 100.0%	The same description with tags stripped.
condition	200/200 100.0%	200/200 100.0%	Condition report as served: HTML, native language.
conditionText	200/200 100.0%	200/200 100.0%	The same condition report with tags stripped.
placement	200/200 100.0%	154/200 77.0%	Where the lot physically sits in the saleroom (shelf/rack code).
categoryId	200/200 100.0%	200/200 100.0%	Auctionet category id.
categoryName	200/200 100.0%	200/200 100.0%	Category leaf name.
categoryPath	200/200 100.0%	200/200 100.0%	Full category path, e.g. Furniture / Armchairs & Chairs.
currency	200/200 100.0%	200/200 100.0%	Currency of every amount on this row — SEK, EUR, GBP or DKK. Never converted.
estimate	200/200 100.0%	200/200 100.0%	The house's estimate (the low estimate where a range exists).
upperEstimate	0/200 0.0%	0/200 0.0%	Upper estimate, on the rare lots that publish a range.
startingBid	200/200 100.0%	200/200 100.0%	Opening bid.
nextBidAmount	200/200 100.0%	200/200 100.0%	What the next valid bid would have to be.
suggestedBidAmounts	200/200 100.0%	200/200 100.0%	The three increments the site offers a bidder.
currentBid	121/200 60.5%	200/200 100.0%	Highest bid so far, or null if nobody has bid.
hammerPrice	0/200 0.0%	200/200 100.0%	Realized price — sold lots only, null on a live lot.
bidCount	200/200 100.0%	200/200 100.0%	Number of bids. A real 0, not a missing value.
bids	121/200 60.5%	200/200 100.0%	The full bid ladder: amount, anonymised bidder number, auto-bid flag, timestamp.
reserveMet	200/200 100.0%	200/200 100.0%	Whether the reserve has been met.
reserveAmount	77/200 38.5%	200/200 100.0%	The reserve, where the house discloses it.
prefersStartingBidToEstimate	200/200 100.0%	200/200 100.0%	The house shows the starting bid rather than an estimate.
estimateRatio	121/200 60.5%	200/200 100.0%	Hammer (or current bid) ÷ estimate. null when either is missing — never a 0.
endsAt	200/200 100.0%	200/200 100.0%	Close time, ISO 8601 UTC.
endsAtUnix	200/200 100.0%	200/200 100.0%	Close time as a unix timestamp.
publishedAt	200/200 100.0%	200/200 100.0%	When the lot went up, ISO 8601 UTC.
publishedAtUnix	200/200 100.0%	200/200 100.0%	Publication time as a unix timestamp.
daysListed	200/200 100.0%	200/200 100.0%	Days between publication and close.
secondsUntilClose	200/200 100.0%	0/200 0.0%	Seconds left to bid. null once the lot has ended.
houseName	200/200 100.0%	200/200 100.0%	Selling auction house.
houseCity	200/200 100.0%	200/200 100.0%	City, from the lot when present and from the house record otherwise.
houseCompanyId	200/200 100.0%	200/200 100.0%	Auctionet company id — the exact join key to the house record.
houseSlug	200/200 100.0%	200/200 100.0%	The house's slug on auctionet.com.
houseUrl	200/200 100.0%	200/200 100.0%	The house's page on auctionet.com.
housePhone	200/200 100.0%	200/200 100.0%	House phone. Neither rival ships this.
houseEmail	200/200 100.0%	200/200 100.0%	House email. Neither rival ships this.
houseWebsite	200/200 100.0%	200/200 100.0%	The house's own website.
houseStreetAddress	200/200 100.0%	200/200 100.0%	Full street address.
housePostalCode	182/200 91.0%	200/200 100.0%	Postcode, parsed out of the address block.
houseCountry	200/200 100.0%	200/200 100.0%	Country.
houseLatitude	200/200 100.0%	200/200 100.0%	Latitude, from the house page's own map link.
houseLongitude	200/200 100.0%	200/200 100.0%	Longitude, from the house page's own map link.
imageUrl	200/200 100.0%	200/200 100.0%	Primary photo, full size.
imageUrlMedium	200/200 100.0%	200/200 100.0%	Primary photo, 640px.
imageUrlThumb	200/200 100.0%	200/200 100.0%	Primary photo, thumbnail.
imageUrls	200/200 100.0%	200/200 100.0%	Every photo, full size.
imageCount	200/200 100.0%	200/200 100.0%	How many photos the lot has.
images	200/200 100.0%	200/200 100.0%	Every photo in all four sizes Auctionet publishes.
sourceUrl	200/200 100.0%	200/200 100.0%	The exact API URL this row came off.
scrapedAt	200/200 100.0%	200/200 100.0%	When this run read the lot.

catalogNumber, eventId and upperEstimate are 0/200 on both pages above, and they are kept anyway — because they are not random gaps, they are catalogued-sale fields. Most Auctionet lots are rolling online sales with a single estimate and no catalogue. The houses that run catalogued sales fill all three on every lot: on fixtures/items_sold_company528_catalogued.json, a real captured page from Ma San Auction, it is 200/200 for all three, with upperEstimate always at or above estimate. Across the wider 8,000-row archive sample it is 3.0% / 3.0% / 3.0%. So: use them when you scrape a catalogue house, and never advertise an estimate RANGE off this source in general — 97% of lots publish one number.

The auction house contact block

Measured over all 86 houses in the directory, every page fetched today (86/86 HTTP 200):

	filled
housePhone	86/86	100.0%
houseEmail	86/86	100.0%
houseWebsite	86/86	100.0%
houseStreetAddress	86/86	100.0%
houseCountry	86/86	100.0%
houseLatitude / houseLongitude	86/86	100.0%
housePostalCode	76/86	88.4%
phone and email on the same house	86/86	100.0%

The join is on the integer company_id that every lot already carries, not on a name match. 100.0% of live lots reach a contact. On the archive it is 99.5%: 5 houses have left Auctionet and their directory pages are gone, taking 17,284 archive lots with them (Leiflers Auktionshus, id 47; Dreweatts Online, id 185; Art and Design Auctions, id 284; Auktionshaus von Brühl, id 75; Auktionshaus Dannenberg, id 52). Those lots still arrive — with housePhone: null — and the run summary names the ids.

A real row

Not an example. This is lot 5239651 exactly as this Actor emitted it on 2026-09-18, trimmed only by cutting repeated bids and images:

{
  "lotId": 5239651,
  "title": "FÅTÖLJ, 1970-tal.",
  "longTitle": "5239651. FÅTÖLJ, 1970-tal.",
  "lotUrl": "https://auctionet.com/en/5239651-armchair-1970s",
  "catalogNumber": null,
  "auctionId": 9332440,
  "eventId": null,
  "lotType": "online",
  "lotStatus": "sold",
  "state": "sold",
  "isEnded": true,
  "isHammerAuctionLot": false,
  "description": "<p>Höjd ca 88 cm\n<br />Sitthöjd ca 40 cm\n<br />Armstöd till armstöd ca 69 cm.\n<br />Skinn, snurrbar fot krom.</p>",
  "descriptionText": "Höjd ca 88 cm\n\nSitthöjd ca 40 cm\n\nArmstöd till armstöd ca 69 cm.\n\nSkinn, snurrbar fot krom.",
  "condition": "<p>Fläckar, repor, mindre skador.</p>",
  "conditionText": "Fläckar, repor, mindre skador.",
  "placement": "H1C",
  "categoryId": 18,
  "categoryName": "Armchairs & Chairs",
  "categoryPath": "Furniture / Armchairs & Chairs",
  "currency": "SEK",
  "estimate": 1200,
  "upperEstimate": null,
  "startingBid": 300,
  "nextBidAmount": 850,
  "suggestedBidAmounts": [
    850,
    900,
    950
  ],
  "currentBid": 800,
  "hammerPrice": 800,
  "bidCount": 4,
  "bids": [
    {
      "bidId": 42739950,
      "bidderNumber": 3,
      "amount": 800,
      "isAutoBid": false,
      "reserveMet": true,
      "placedAt": "2026-09-18T19:07:16.000Z"
    },
    {
      "bidId": 42739881,
      "bidderNumber": 3,
      "amount": 400,
      "isAutoBid": false,
      "reserveMet": false,
      "placedAt": "2026-09-18T19:05:37.000Z"
    },
    … 2 more bids
  ],
  "reserveMet": true,
  "reserveAmount": 800,
  "prefersStartingBidToEstimate": false,
  "estimateRatio": 0.6667,
  "endsAt": "2026-09-18T19:09:00.000Z",
  "endsAtUnix": 1789758540,
  "publishedAt": "2026-09-11T12:58:40.000Z",
  "publishedAtUnix": 1789131520,
  "daysListed": 7.26,
  "secondsUntilClose": null,
  "houseName": "Gomér & Andersson Linköping",
  "houseCity": "Linköping",
  "houseCompanyId": 9,
  "houseSlug": "ga",
  "houseUrl": "https://auctionet.com/en/ga",
  "housePhone": "+46 (0)13-375210",
  "houseEmail": "linkoping@gomer-andersson.se",
  "houseWebsite": "http://www.gomer-andersson.se/",
  "houseStreetAddress": "Industrigatan 36, 582 77 Linköping, Sweden",
  "housePostalCode": "582 77",
  "houseCountry": "Sweden",
  "houseLatitude": 58.419569,
  "houseLongitude": 15.606969,
  "imageUrl": "https://images.auctionet.com/thumbs/hd_item_5239651_fba399e43a.jpg",
  "imageUrlMedium": "https://images.auctionet.com/thumbs/w640_item_5239651_fba399e43a.jpg",
  "imageUrlThumb": "https://images.auctionet.com/thumbs/thumb_item_5239651_fba399e43a.jpg",
  "imageUrls": [
    "https://images.auctionet.com/thumbs/hd_item_5239651_fba399e43a.jpg",
    "https://images.auctionet.com/thumbs/hd_item_5239651_30c573550b.jpg",
    … 2 more
  ],
  "imageCount": 7,
  "images": [
    {
      "thumb": "https://images.auctionet.com/thumbs/thumb_item_5239651_fba399e43a.jpg",
      "mini": "https://images.auctionet.com/thumbs/mini_item_5239651_fba399e43a.jpg",
      "medium": "https://images.auctionet.com/thumbs/w640_item_5239651_fba399e43a.jpg",
      "full": "https://images.auctionet.com/thumbs/hd_item_5239651_fba399e43a.jpg"
    },
    … 3 more
  ],
  "sourceUrl": "https://auctionet.com/api/v2/items.json?per_page=200&page=1&company_id=9&is=ended",
  "scrapedAt": "2026-09-18T19:09:59.670Z"
}

Inputs

Input	What it does
Live lots or the sold archive	live (default) or sold.
Search term	Free text across titles and descriptions. Descriptions are in each house's own language, so a Swedish word finds more Swedish lots than its English translation.
Auction houses	Pick from all 86, each shown with its live and sold lot counts.
Extra auction house IDs	Numeric company_id, for the 5 departed houses that still have archive lots.
Categories	Auctionet's own tree, all 160 nodes with counts. Picking a top-level category includes everything under it.
Countries	SE, DK, FI, DE, ES, GB.
Minimum / maximum estimate	In the lot's own currency.
Only lots that already have a bid	121/200 on the captured ending-soonest page — but 3,797/8,000 (47.5%) across the wider live sample, which is the honest corpus-wide number.
Sort order	Ending soonest, most recent, highest/lowest estimate, highest/lowest bid, most/fewest bids.
Join the house's contact details	On by default.
Include the full bid ladder	On by default. Sold lots average 7.05 bids, live lots 3.32.
Include every image URL	On by default. Every lot in the sample has a photo — 5.44 on average live, 6.70 sold.
Max lots	Hard cap on rows delivered and charged. Default 1,000.
Proxy	Apify Proxy, datacenter. Verified below.

Every value you can type is checked against the live site before a single row is delivered, and a bad one fails the run with the valid list rather than charging you. That is not defensive politeness — this API answers HTTP 200 to a filter it does not recognise and returns the unfiltered catalogue: category_id=999999 returns all 39,189 live lots (captured in fixtures/ignored_param_category_999999.json), an unknown company_id does the same, and country_code=UK returns 0 rows that look exactly like a real empty result. One typo would otherwise be a bill for the whole store.

How it gets past the source's 10,000-lot ceiling

This is the one thing about auctionet.com that is not in any documentation, and the reason a naive scraper cannot give you the archive.

Every distinct query serves at most 10,000 items. At per_page=200 (the ceiling; asking for 500 or 1,000 silently returns 200), page 50 returns 200 lots, page 51 returns an 88-byte body with zero lots, and page 60 returns HTTP 302 back to page 1. pagination.total_pages caps itself at 10,000 / per_page, so a 247,488-lot bucket reports 50 pages, not 1,238. A page-walker sized off total_entries therefore stops dead at 10,000 rows and reports success. Both boundary responses are in fixtures/window_perpage20_page500.json and …page501.json.

So this Actor partitions until every query fits inside the window, preferring the axis measured lossless:

1. Category — the 25 top-level categories sum to 39,189 of 39,189 live lots and 3,938,923 of 3,938,923 sold lots — exactly 100.00% of both — and a parent query includes its children, so the tree can be descended until each node fits.
2. Auction house — read free off the search page's own facet panel, which returns every company_id with its exact lot count for that query, so a house split costs one request and no probing. 84 houses cover 100% of live lots; 91 cover 89.1% of the archive.
3. Country — the same facet panel, six values.
4. Sort order — not a partition but a window multiplier: a different sort opens a different 10,000-lot window over the same bucket. Measured pairwise on a >10k bucket: end_asc and end_desc are not opposites (104/200 ids in common on page 1), but estimate_asc/estimate_desc/bid_asc/bid_desc/bids_count_asc/bids_count_desc are disjoint (0/200 overlap), so seven orderings reach up to 70,000 lots in one bucket.

Rows are deduplicated by lot id, so an overlap costs requests and never a duplicate row. Whatever no axis can reach is counted and reported in RUN_SUMMARY.coverage, not quietly dropped.

Two more things the partitioner does that save you money:

• It only splits when the window is actually in the way. Asking for 40 lots out of a 3.9M-lot archive needs no partitioning at all, so the demo run costs ~11 API calls rather than 25 category probes.
• An estimate filter cannot be pushed to the source (min_estimate, estimate_from and friends are all silently ignored), so instead of scanning and discarding, the Actor pins the estimate sort and walks only the matching end. Measured on one house's archive with minEstimate: 5000, maxItems: 60: 200 lots read instead of 5,357, one page instead of 27, 1.2 MB instead of 13.2 MB — same 60 rows delivered.

Billing

$3.00 per 1,000 lots delivered. No run-start fee. One charge event, lot-scraped, billed per row actually written to your dataset.

• Rows removed by your filters are never delivered and never charged.
• Max lots is checked against rows produced, not batches flushed — a cap of 37 delivers and charges exactly 37. The validator proves this at caps of 1, 37, 199, 200, 201 and 350.
• A run that matches nothing costs nothing. So does a run that fails validation.
• The house contact join, the bid ladder, the hammer price and the category path are all included in the same rate — there is no enrichment upsell.

Honest limits

1. The archive contact join tops out at 99.5%. 5 auction houses have left Auctionet; their lots stay in the archive but their directory pages are gone, so 17,284 sold lots arrive with housePhone: null. The run summary lists the ids.
2. Sold lots only in the archive. is=ended returns sold lots — 200/200 in the captured page are state: "sold" with a bid on every one. Auctionet does not publish unsold/passed lots, so sell-through rate is not derivable from this source. Anyone who tells you otherwise is inferring it.
3. There is no hammer_price field. hammerPrice is max(bids[].amount) on a sold lot. It is cross-checked against the lot page's own embedded record in the validator: for lot 5239651 the API says 800 SEK against an estimate of 1200, and the page — which renders in the viewer's currency, not the house's — shows the same ratio to within 0.02%.
4. hammered does not mean "sold". Auctionet's own flag marks an in-room hammer auction lot; it was false on 200/200 lots of a page of sold archive rows. It ships as isHammerAuctionLot so its name cannot mislead you. Sold-ness comes from state.
5. Amounts are never converted. Every figure is in the lot's currency (SEK 347, EUR 28, GBP 18, DKK 7 across the 400 sample rows). Comparing a SEK hammer price to a GBP one is your conversion to do.
6. Descriptions and conditions are HTML in the house's own language — Swedish, Danish, German or Spanish. They are passed through verbatim plus a plain-text twin. Nothing is translated, because a guessed translation yields a plausible wrong answer rather than an error.
7. upperEstimate, catalogNumber and eventId are catalogue-sale fields, not general ones (3.0%, 3.0% and 3.0% of 8,000 archive rows overall — but 200/200 at a house that runs catalogued sales). Filter to those houses if you need them; do not build a range-based model on the corpus as a whole.
8. A capped run is a prefix, not a sample. Buckets are walked depth-first, so maxItems: 500 returns whole categories and houses rather than 500 lots spread evenly across the corpus. Use the filters when you want a specific slice.
9. Live lots drift while you read them. total_entries moved from 39,548 to 39,047 over three hours of measurement today as lots closed. A 4% cross-page id overlap under a shifting sort is normal and is deduplicated by id.
10. bidderNumber is a paddle number, not a person. Auctionet publishes an anonymous per-lot bidder index (1, 2, 3…) on the public lot page. No account, name or identity is available, and none is collected.

Source and legal

• API: https://auctionet.com/api/v2/items.json — public, unauthenticated, CORS-open (access-control-allow-origin: *), and the live site's own data source: the lot page's data-react-props carries the same record.
• robots.txt, captured verbatim in fixtures/robots.txt:

  User-Agent: *
  Disallow: /admin/
  Disallow: /*/my$
  Disallow: /*/my/*
  Sitemap: https://auctionet-exports.s3-eu-west-1.amazonaws.com/sitemap.xml.gz
  

  Neither /api/ nor the lot pages nor the auction-house pages are disallowed.
• Only public listing and business-contact data is read. No login, no personal data, no bidder identities.
• Auctionet is a trademark of its owner. This is an independent tool and is not affiliated with, endorsed by or connected to Auctionet.

Transport, measured

Cloudflare challenge	none — the API returns HTTP 200 to a request with no User-Agent header at all
Apify Proxy, datacenter	HTTP 200; a 842,986-byte body arrived complete, Content-Length matched
Apify Proxy, RESIDENTIAL	HTTP 200, same
Sustained datacenter load	12 large bodies through fresh sessions: 11 complete, 0 truncated, 1 socket hang-up (retried on a fresh session)
Throughput	60 requests at concurrency 10 → 60/60 HTTP 200, 9,792 lots, 12.6 s
1,000 lots delivered	89 requests, 3.4 MB, 13 s (live) / 90 requests, 4.9 MB, 13 s (sold)
Peak memory on a 3,000-lot run	225 MB, which is why this Actor is configured at 512 MB

Large bodies are checked against Content-Length before parsing, because a multi-hundred-KB body can truncate mid-stream on a datacenter proxy while still reporting HTTP 200. A non-200, or a 200 carrying a challenge page, is treated as a transport failure and retried on a fresh proxy session — never as "this query is empty". If a large fraction of requests fail, the run fails rather than reporting a partial read as a complete one.

What you get that the alternatives do not

Two other Actors scrape this site. Both were read in full on 2026-09-18. Both charge Apify's apify-actor-start event as well as a per-result rate, and that start fee is billed once per gigabyte of allocated memory — 2 GB and $0.01 for one, 4 GB and $0.02 for the other.

	this Actor	stealth_mode/auctionet-product-search-scraper	crawlerbros/auctionet-scraper
Price (FREE tier)	$0.003/lot, no start fee	$0.01/result + a start fee	$0.005/result + a start fee
A 200-lot run costs	$0.60	$2.01	$1.02
Fields per row	61	19	16
Lot description + condition	✅ raw and plain text	❌	✅ (stripped only)
Category path	✅ from the live 160-node tree	❌	✅
Estimate / starting bid / next bid	✅	estimate only	estimate only
Current bid / bid count	✅	display string only	✅
Full bid ladder with timestamps	✅	❌	❌
Realized hammer price	✅ 100% of archive rows	❌	via a keyword-only "ended" mode
Hammer ÷ estimate ratio	✅	❌	❌
Auction house phone + email	✅ 100.0%	❌	❌
House address + coordinates	✅ 100.0%	❌	❌
Reserve amount / reserve met	✅	reserve met only	reserve met only
Images	✅ all 4 sizes	✅ 1 size	✅ 3 sizes
Filter by house / category / country	✅ validated against the live site	✅ (unvalidated)	category only
Gets past the 10,000-per-query ceiling	✅ partitioned, with coverage reported	❌	❌ (README caps a run at 5,000)
Run summary reconciling source total vs delivered	✅	❌	❌
Per-field fill measured and published	✅	❌	❌

Fields the rivals ship that this Actor cuts, and why:

• highlighted, canBeFollowed, isFollowing, showAuctionEndTime — viewer/session UI state, not lot data. highlighted was false on 48/48 cards in a real captured search page.
• amountLabel, amountValue, amountTitle — display strings converted into the viewer's session currency ("201 EUR"). This Actor ships the numbers in the lot's own currency instead, plus the currency code, so you can convert with a rate you control.
• shortTitle — the truncated card title. title is the full one, and longTitle is the site's "<id>. <title>" form.
• eventTitle — the themed-sale name. It is only on the search card, not the API, and it was populated on 1/48 cards; eventId ships instead.
• scrapedAt (crawlerbros) is not cut — it ships, along with sourceUrl, the exact API URL each row came off.

Run summary

Every run writes RUN_SUMMARY to the key-value store: lots delivered (which is exactly lots charged), lots read, rows dropped by your filters, duplicate ids skipped, how the query was partitioned and anything no axis could reach, per-field fill counted on your own rows, house-contact coverage with the company ids that no longer have a directory page, currency and house breakdowns, and every URL that failed after retries. On a failed run it carries the reason and the failing URLs, so an unreachable source is never mistaken for an empty one.

Validating this build

node offline_validate.mjs

No network, no platform, no token. It parses the captured bytes in fixtures/ with the same modules the Actor runs, and asserts every number in this README.