🩺 MEDICA & Messe Düsseldorf Exhibitor Scraper — filter by product category, not just "download the list"

Messe Düsseldorf runs some of the largest B2B trade fairs on earth — MEDICA, COMPAMED, K, ProWein, interpack, drupa, EuroShop, A+A, boot, glasstec and more. Every one of them publishes a searchable exhibitor directory. This Actor turns any of them into a clean company table: name, country, city, postcode, hall and stand, and — optionally — phone, website, full street address, profile description, product list and named contact people.

The part that makes it a targeting tool rather than a dump: the fairs classify their own exhibitors into a product-category tree, and this Actor lets you scrape a single branch of it. A lab-diagnostics distributor does not want all 3,576 MEDICA exhibitors. They want the 515 in Laboratory, IVD & Diagnostics — and the 64 of those that are German. Both numbers below are real, from real runs, and both are one input field away.

---

📊 What you get, per row

Every row carries the same key set. A listing-only run emits 27 columns; switching on Fetch full detail adds 14, and contact persons add 2 more (41 and 43 columns, verified on runs JGMPcspdZCLhHaq2L, x72sdbMHNw8KorvjT and BfilxVkncm1Ut2ZBl — 1 key set in each).

The listing scrape — what every run returns

Fill measured across the entire MEDICA + COMPAMED corpus, all 3,576 rows, run ZSzg00nzcPc5MCt6e (2026-09-08, build 0.1.2, complete sweep, 0 duplicates):

Field	Fill	What it is
exhibitorId	100%	The fair's own id (medica2026.3081101) — the stable de-duplication key
name	100%	🏢 Company name, exactly as the fair publishes it
country / city	100%	Where the company is based (not where its stand is)
postalCode	99.8%	Company postcode
hallStand	100%	📍 Hall 8b / J21 — the whole booth string
hall / stand / halls[]	100%	The same thing split, plus every extra booth a company holds
eventId	100%	Which fair inside a shared directory (medica2026 vs compamed2026)
eventName	100% on shared directories	The event's display label. Fairs whose directory holds only one event publish no per-row label, so this is null there — use fairName and eventId (see FAQ)
fairDomain / fairName	100%	Which fair site this row came from
fairStartDate / fairEndDate	100%	The edition's dates, read live from the fair's own config — so you always know which edition you scraped
profileUrl	100%	Deep link to the exhibitor's public profile page
isPremiumExhibitor	100%	Whether the fair flags them as a premium exhibitor
teaser	16.6%	The short blurb shown on the search card (92% on K 2025 — see the maturity note)
tags[]	14.9%	Self-declared criteria: Manufacturer · OEM · Service · Premium Exhibitor (MEDICA 2026, measured 2026-09-08: 477 / 193 / 191 / 5 — these move as exhibitors register)
logoUrl	10.4%	Company logo / stand image
queryLabel / queryDeclaredTotal	100%	Which of your filters produced the row, and how many the fair said that filter has — so you can always see whether you got all of them
categoryId / categoryLabel / categoryPath	100% when you filter by category	e.g. medica2026.01.02 · Laboratory, IVD & Diagnostics · MEDICA 2026 > Clinical Diagnostics & Therapeutic Technologies > Laboratory, IVD & Diagnostics
scrapedAt	100%	ISO timestamp

🔎 With Fetch full detail switched on

One extra request per exhibitor. Measured on two real platform runs, because the numbers differ a lot by how close the fair is (see the maturity note under it):

Field	MEDICA 2026 (n=50, run x72sdbMHNw8KorvjT)	K 2025 (n=150, run BfilxVkncm1Ut2ZBl)	What it is
phone	100%	99%	☎️ Company switchboard, international format
website	94%	99%	🌐 Company website — the field you pipe into an e-mail-finding tool
street	100%	100%	Street and number
addressPostalCode / addressLocality	98% / 100%	99% / 100%	Postcode and town from the detail record
countryCode	100%	100%	ISO-2 (DE, IT, CN)
addressRegion	24%	18%	State / region where the exhibitor gave one
poBox	6%	7%	Post-office box
description	30%	92%	The company's own profile text, HTML stripped to clean text (or kept as HTML — your choice)
products[]	16%	86%	🧪 The product names the exhibitor lists on their own profile
brands[]	4%	3%	Registered brands
socialProfiles[]	0% / 0%	0%	Company-level social links (present in the schema, rarely filled by exhibitors)
emailsInDescription[]	0%	1%	⚠️ Not an e-mail field — read the honest-limits section. These are addresses an exhibitor happened to type into their own description text
detailFetched	100%	100%	Whether the detail request actually succeeded for this row

📅 The maturity effect — the single most important thing to understand about fill rates. MEDICA 2026 opens on 2026-11-16. Its exhibitors are still filling in their profiles, so descriptions sit at 30% today and will climb. A fair that has already happened is far richer: K 2025 description 92%, products 86%. The address block does not move — phone, street and town are ~100% on both, because the fair collects those at registration. If you need rich descriptions right now, scrape a completed fair (K, A+A, drupa, GMTN).

👤 With Include contact persons switched on

The named staff an exhibitor chose to publish on its own profile — name, givenName, familyName, jobTitle, phone, linkedInUrl, plus a contactCount per company.

Fair	Exhibitors with ≥1 contact	People returned	Most on one company
K 2025 (n=150, run BfilxVkncm1Ut2ZBl)	59%	197	9
MEDICA 2026 (n=30)	40%	23	5
ProWein 2026 (n=30)	50%	17	2
A+A 2025 (n=30)	40%	26	6

Off by default, priced separately, and covered by its own paragraph in Privacy below.

---

🎪 Which fairs — 17 hostnames, 15 distinct directories, 28,698 exhibitors

Every hostname below was probed live on 2026-09-07 and answered with a real exhibitor list; the count is what the fair itself reported that day. Pick one from the dropdown, or type any other Messe Düsseldorf hostname into Custom fair hostname — it is validated before anything is scraped.

Fair	Exhibitors	Notes
MEDICA — medical technology	3,576	shares its directory with COMPAMED (MEDICA 2,842 + COMPAMED 734)
COMPAMED — medical components	3,576	same directory; split with the Events filter
K — plastics & rubber	3,237
ProWein — wine & spirits	3,105	adds a sales-channel filter
interpack — packaging & processing	2,932	shares with components (2,825 + 107)
wire — wire & cable	2,656	shares with Tube (1,510 + 1,146)
Tube — tube & pipe	2,656	same directory as wire
A+A — safety & health at work	2,364	completed fair, rich profiles
GIFA / METEC / THERMPROCESS / NEWCAST	2,205	four fairs in one directory (881 / 570 / 280 / 487)
EuroShop — retail	1,900
drupa — print technology	1,630	2024 edition, still served in full
boot Düsseldorf — water sports	1,467
glasstec — glass	1,173
CARAVAN SALON — caravanning	873
Valve World Expo — valves	549
EuroCIS — retail technology	523
BEAUTY DÜSSELDORF — cosmetics	508

One run covers one fair. To sweep several, run the Actor once per fair (or schedule a task per fair). Some Messe Düsseldorf brands are not on this platform at all — REIFEN, TOP HAIR, Energy Storage Europe and a few others answer "No config set found"; the Actor stops with that exact message and lists the ones that do work, instead of returning the wrong fair.

---

🎯 Who buys this

• Distributors and importers scouting manufacturers in one product niche — the category filter is the whole point: Laboratory, IVD & Diagnostics is 515 companies, not 3,576.
• Contract manufacturers, packaging and sterile-barrier suppliers selling into the exhibitor base of MEDICA/COMPAMED or interpack.
• Trade-show and event agencies building booth-services, logistics and stand-build prospect lists — filter by hall and you have the physical neighbourhood.
• Market researchers mapping who exhibits where, by country, over multiple editions (drupa 2024 and GMTN 2023 are still served in full).
• Sales teams working a fair: sort by Newest registered first to see who signed up most recently, and by hall and stand to plan the walk.

---

🚀 Example inputs

1 — a whole fair, cheapest possible (listing only)

{ "fair": "www.medica-tradefair.com", "maxItems": 0 }

3,576 rows, 37.7 s, $3.58. Verified: run ZSzg00nzcPc5MCt6e, 3,576 delivered / 3,576 charged.

2 — one product category, with full contact detail

{
  "fair": "www.medica-tradefair.com",
  "productCategories": ["Laboratory, IVD & Diagnostics"],
  "includeDetail": true,
  "maxItems": 0
}

515 rows. The listing-only version of this run (uHXQwOu4neWhyZiqB) took 19.7 s and matched the fair's own published count for that category exactly: 515 declared, 515 delivered, 0 duplicates.

3 — German lab-diagnostics manufacturers only

{
  "fair": "www.medica-tradefair.com",
  "productCategories": ["Laboratory, IVD & Diagnostics"],
  "countries": ["DE"],
  "exhibitorCriteria": ["Manufacturer"],
  "includeDetail": true,
  "maxItems": 0
}

The category × country pair alone is 64 exhibitors of 3,576.

4 — everyone in one hall, for booth services

{ "fair": "www.medica-tradefair.com", "halls": ["Hall 1"], "sortBy": "location_asc", "maxItems": 0 }

5 — keyword hunt across a fair

{ "fair": "www.medica-tradefair.com", "searchTerms": ["dialysis", "hemodialysis", "endoscope"], "maxItems": 0 }

Each keyword is its own query and results are de-duplicated: measured 8 + 1 + 4 = 13 hits, 10 unique companies delivered, 3 duplicates suppressed and not charged.

6 — paste a URL from the fair's own website

{ "startUrls": [{ "url": "https://www.k-online.com/vis/v1/en/search?_query=extruder&f_country=DE" }] }

The host picks the fair, and the URL's filters and keyword are carried into the run.

7 — one company, everything about it

{
  "startUrls": [{ "url": "https://www.medica-tradefair.com/vis/v1/en/exhprofiles/LBwn2IUJTPWeGs2rx27Jaw" }],
  "includeDetail": true,
  "includeContactPersons": true
}

8 — the German edition of a fair

{ "fair": "www.k-online.com", "language": "de", "includeDetail": true, "maxItems": 0 }

Country names, hall labels, category names and the profile description all come back in German — exhibitors often write the two language versions separately.

---

🔎 What you can filter on

All of it runs server-side at the fair, so you are only billed for rows you asked for.

Filter	Accepts	Measured example (MEDICA 2026)
Product categories	the category name or its id	68 categories over 3 levels; Laboratory, IVD & Diagnostics = 515
Countries	ISO-2 or country name	Germany 644 · China 688 · France 144 (70 countries listed)
Halls	hall id or label	18 hall areas; Hall 1 = 245
Events	event name or id	MEDICA 2026 = 2,842 · COMPAMED 2026 = 734
Exhibitor criteria	tag label or id	Manufacturer · OEM · Service (MEDICA 2026 counts move daily; 477 / 193 / 191 on 2026-09-08)
Keywords	free text	dialysis 8 · endoscope 15 · sterile packaging 16 · ultrasound 44
Other filters	f_targrp=…, f_industry=…, f_sales=…	fair-specific: K and drupa have target groups, ProWein a sales channel, EuroShop its own tags
Sort	A-Z · Z-A · newest registered · oldest · by hall · relevance	changes which rows you get when maxItems cuts the sweep short

Every filter value is checked against the fair's live filter list before the crawl starts. Type "Wakanda" as a country and the run stops and prints the 70 countries that fair actually has — it does not quietly drop the filter and bill you for all 3,576 rows. That failure is a real, verified behaviour: run KbqjOhhxEkjPqMaGl, FAILED in 10 s, $0.00 charged. The full live vocabulary for the fair you picked is also written to the run's key-value store as FAIR_FACETS, so the next run can use exact ids.

---

💵 Pricing

Pay per event. No monthly fee, no charge for starting a run, and no charge for a run that returns nothing.

Event	Price	When it fires
Per exhibitor returned	$0.001	once per company written to your dataset
Per detail fetch	$0.002	only with Fetch full detail on, and only for rows that were actually delivered
Per contact person	$0.001	only with Include contact persons on, per named person returned

What you want	Cost
All 3,576 MEDICA + COMPAMED exhibitors, listing only	$3.58
The same, with phone / website / street address	$10.73
One product category (515 exhibitors) with full detail	$1.55
All 15 directories, 28,698 exhibitors, listing only	$28.70
A 50-row trial with full detail	$0.15

How this compares. The established Messe Düsseldorf exhibitor scraper on this Store charges $0.005 per exhibitor (free tier), $0.004 on paid tiers, plus a $0.01 fee for starting each run, and runs at 1024 MB. Our listing row is 5× cheaper, our full-detail row is 25% cheaper, there is no start fee, and this Actor runs at 256 MB — measured peak memory use was 64 MB on the full 3,576-row sweep.

Billing integrity, and how to check it yourself. Every run above was verified with chargedEventCounts equal to rows delivered, after waiting for Apify's counters to settle:

Run	Rows delivered	exhibitor-scraped	exhibitor-detail-enriched	contact-person-scraped
ZSzg00nzcPc5MCt6e (full sweep)	3,576	3,576	0	0
JGMPcspdZCLhHaq2L (defaults)	250	250	0	0
x72sdbMHNw8KorvjT (50 + detail)	50	50	50	0
BfilxVkncm1Ut2ZBl (150 + detail + contacts)	150	150	150	197
uHXQwOu4neWhyZiqB (category, full)	515	515	0	0
KbqjOhhxEkjPqMaGl (bad filter)	0	0	0	0

Duplicates are removed before billing, filtered-out rows are never charged, and if you set a maxTotalChargeUsd cap the run trims the batch before it spends a detail request — so you never receive enrichment your cap cannot pay for, and never pay for enrichment you did not receive.

---

⚙️ How it works, and what it does not hide

Messe Düsseldorf serves every fair's exhibitor finder from one platform, and the fair is selected by a single request header. This Actor talks to that platform's own public JSON API directly — no browser, no login, no CAPTCHA, no cookies — which is why it runs at 256 MB and finishes a 3,576-company fair in under 40 seconds.

Transport, measured through Apify's automatic (datacenter) proxy: 297 of 300 detail requests succeeded first try, the other 3 were connection resets recovered by retry; 13 requests/second at concurrency 10; median response 622 ms, 95th percentile 1.24 s. Residential proxies also work and are not needed. The platform does rate-limit a single address pushed hard, so HTTP 429 is retried with the server's own Retry-After.

Three things this Actor refuses to paper over:

1. A filter the platform accepts but silently ignores. The product-category tree exists in two different id-spaces, and passing the wrong one returns the entire fair with a 200 OK. Before delivering anything, the Actor compares the result count against the count the platform itself published for that exact filter. If a filter did not filter, the run fails and delivers nothing rather than billing you for thousands of rows you did not ask for.
2. A truncated sweep is never reported as a complete one. Every row carries queryDeclaredTotal — the fair's own count for the filter that produced it — and the run's status message names whatever cut the sweep short (maxItems, the run timeout, or your charge cap). The run's key-value store holds a RUN_SUMMARY record with a per-query complete flag and cutShortBy reason.
3. A zero-row run says why. "No exhibitor matched" is only ever printed when the platform was actually asked and answered. A run that ran out of time, hit a charge cap, or failed at the transport layer says exactly that instead, and a transport failure ends the run FAILED rather than pretending the fair is empty.

---

🚫 Honest limits — read before you buy

• There is no e-mail address in this data. The platform simply does not publish one. We checked hard: 300 exhibitor detail payloads across 4 different fairs were regex-scanned in full, and 4 of them (1.3%) contained an e-mail-shaped string — every one of those inside free text the exhibitor typed into their own description. Those are surfaced honestly in emailsInDescription[] and nowhere else. What you do get is website at 94–99% and phone at 99–100%, which is what an e-mail-finding or contact-page tool needs as input. If a listing promises you exhibitor e-mails from this source, be sceptical.
• Contact persons are not on every exhibitor — 40% on MEDICA 2026, 59% on K 2025. And they are named individuals: see Privacy.
• Description and product lists are thin on a fair that has not happened yet (MEDICA 2026: 30% and 16% today) and rich on one that has (K 2025: 92% and 86%). This is a calendar fact about the source, not a limitation of the scraper.
• eventName is null on single-fair directories. The platform only publishes a per-row event label where a directory holds more than one fair. fairName and eventId are always filled.
• One run, one fair. There is no "all fairs" mode, because there is no route on the platform that lists the fairs — the 17 hostnames are hand-verified, and a new one can be typed into Custom fair hostname without waiting for an update.
• Fairs are biennial or triennial, so editions roll over (K 2025 → K 2027, drupa 2024 → 2028). Every row carries fairStartDate, fairEndDate and eventId read live from the fair's own config, so you always know exactly which edition you scraped.
• Ratings, visitor numbers, stand size and floor-plan coordinates are not published by this platform and are not returned.

---

🔒 Privacy — read this before switching contacts on

includeContactPersons returns named natural persons — first name, surname, job title, business phone and LinkedIn URL — that exhibiting companies chose to publish on their own public fair profiles. Most of them are in the EU, so the GDPR applies to you as the controller of that copy. It is off by default, priced as its own event so that turning it on is a deliberate act, and it is the only part of this Actor that touches personal data.

You need a lawful basis (in B2B prospecting this is usually legitimate interest), and you owe those people the transparency, objection and erasure rights that come with it. Everything else the Actor returns — company name, address, switchboard, stand, categories — is ordinary company data. This paragraph is a description of the data, not legal advice.

---

⚖️ Legal and fair use

This Actor reads only publicly available exhibitor directory pages — the same pages any visitor sees without logging in. No login, no paywall, no personal account, and no CAPTCHA is circumvented.

https://finder.messe-duesseldorf.de/robots.txt, in full, on 2026-09-07:

User-agent: *
Disallow:  /kati-cgi/kati/

That is the entire file. The one disallowed path is a legacy CGI directory; none of the routes this Actor uses fall under it. Each fair site's own robots.txt carries the same single Disallow line, followed by its sitemap list — on www.medica-tradefair.com one of those sitemaps is labelled Exhibitors & Products, i.e. the fair publishes these very pages for search engines to crawl. Exhibitor profile pages additionally carry <meta name="robots" content="index,follow">.

You are responsible for complying with Messe Düsseldorf's terms of use and with data-protection law in your jurisdiction, especially if you enable contact persons. This Actor is not affiliated with, endorsed by or connected to Messe Düsseldorf GmbH; MEDICA, COMPAMED, K, drupa, interpack, ProWein, EuroShop, A+A, boot, glasstec, GIFA, METEC, NEWCAST, THERMPROCESS, EuroCIS, BEAUTY and CARAVAN SALON are their trade marks, used here only to say which public directories are supported.

---

❓ FAQ

Which fairs are supported? 17 fair hostnames covering 15 distinct exhibitor directories and 28,698 exhibitors — the table above lists every one with its live count. Any other Messe Düsseldorf hostname can be typed into Custom fair hostname; it is validated against the platform first.

Can I scrape several fairs in one run? No — one run covers one fair, which keeps billing and de-duplication honest. Create one Apify task per fair and schedule them; each takes well under a minute for a listing-only sweep.

How do I find the exact product-category name? Type the name as you see it on the fair's own site — matching is case-insensitive and accepts a distinctive fragment. Or run once with no category and read FAIR_FACETS from the run's key-value store: it lists every filter id, label and live count that fair accepts, including the top of the category tree. Category ids like medica2026.01.02 also work directly.

Why did my run stop with "filter value could not be applied"? Because applying no filter would have returned — and billed you for — the whole fair. The error lists the values that fair really has. Copy one and re-run; nothing was charged.

Does it get e-mail addresses? No, and nothing on this page implies otherwise. The platform publishes website and phone, not e-mail. See Honest limits for the 300-profile measurement behind that statement.

How fresh is the data? It is read live at run time; there is no cache. Exhibitor lists for an upcoming fair change weekly as companies register — sort by Newest registered first to see the new arrivals.

Can I get exhibitors from a fair that already happened? Yes. drupa 2024 still serves 1,630 exhibitors and the GMTN 2023 group serves 2,205. Completed fairs also have the richest profiles.

What is the queryDeclaredTotal column for? It is the fair's own count for the filter that produced that row. Compare it with how many rows you got: equal means you have the complete set, fewer means something truncated the sweep — and the run status message and RUN_SUMMARY name what.

Will a run ever bill me for rows I did not want? The design says no, in four places: filters are validated against the live vocabulary before the crawl; a filter that does not actually filter fails the run; duplicates are removed before billing; and the second-order events (detail, contacts) are charged only for rows that were actually delivered.

What if I set a maximum charge for the run? The run reserves money, not row counts. It trims the batch before spending detail requests, tells you in the log and status message that the cap was what stopped it, and never delivers a row it could not bill for.

Why 256 MB when other exhibitor scrapers use 1024 MB? Because this one uses no browser. Peak memory on the full 3,576-row sweep was 64 MB. You pay less platform compute for the same rows.

Can I get the raw data behind a field you did not map? Yes — switch on Include the raw source payload and each row carries the untouched listing document and the exhibitor's schema.org JSON-LD.

Does it work with the Apify API, schedules and integrations? Yes — it is a normal Actor. Results go to a dataset you can pull as JSON, CSV, Excel or XML, and the dataset ships with four ready-made views (Exhibitors, Company leads, Contact persons, By product category).