226K Listings + 3.4m Images from Immobilienscout24

The client needed the residential side of Immobilienscout24.de in full: not a sample, not a filtered slice, but every listing the portal publishes, together with its photos. Two parts of the brief shaped the build. The scope covered both sales and rentals across two separate markets, and the images were not an afterthought - they had to end up on the client's own server, resized for uniformity.

The data fed an AI-based property search product, and the first complete version had a fixed demo date in early June 2024.

Sources: immobilienscout24.de
Coverage: Buy and Rent categories, residential property in Germany and Mallorca (Spain); offices in Germany added in 2025
Frequency: three full runs on repeat orders: spring 2024, November-December 2024, and April-June 2025
Volume: 226,000 properties and 3.4 million resized images; 420 GB of resized photos on the client's server by April 2025
Delivery: full listing details in CSV files, plus all images downloaded to the client's server as WebP and resized for uniformity, in folders by market, property type, deal type, and date
Setup and launch: 4 working days

Complete Immobilienscout24 listings, not a shortlist of fields

The request was for all property details, not for a set of fields agreed in advance. That changes how the scraper is written. Instead of targeting a handful of known elements, it takes what a listing page actually publishes, so information is not silently dropped when a property carries something the narrower template would have ignored.

Coverage worked the same way. Buy and Rent are separate inventories, and Germany and Mallorca are separate markets with separate volumes. All four combinations had to be walked to completion and merged into one consistent dataset rather than four loosely related exports.

3.4 million images are their own pipeline

At this scale, collecting text and collecting images are two different jobs. Across 226,000 properties there were 3.4 million photos to fetch, and each had to be transferred to the client's server rather than simply referenced by a link. A URL pointing back at the portal is not an asset the client controls; a file on their own storage is.

Every image went through the same step on the way in, set by the client's own standard: the longest side up to 1,500 px, smaller images never enlarged, WebP at quality 75. Handling this during collection means nobody has to make a second pass over millions of files. By April 2025 the resized photos on the client's server took 420 GB.

Four working days from brief to running scraper

Setup and launch took four working days. That window covered working out how the source is structured, building the collection logic for both categories in both markets, and wiring the image download and resize path through to the client's server. The size of the dataset is a matter of run time; the build itself is short because the work is routine for us.

Three seasons of Immobilienscout24 data on repeat orders

The client came back twice for a fresh full export. The second run started on 12 November 2024 and had all of Germany - houses and apartments, buy and rent - on the client's storage by 2 December; German houses for sale alone brought 1.65 million images. The third ran from April to June 2025 and added offices for sale and for rent in Germany. Mallorca has no office section on the portal.

Speed mattered from the first season. In May 2024 the portal changed its bot protection, moving between a browser challenge and a slider captcha. We switched the collection to a lighter route that does not render every page and ran at about 720 listings a minute, roughly seven times the 100 a minute the client's own scraper had managed.

What the client ended up with

A dataset of 226,000 residential properties from Immobilienscout24.de, covering the Buy and Rent categories in Germany and Mallorca, with the full details of each listing. Alongside it, 3.4 million images on the client's own infrastructure, all resized for uniformity and tied to the listings they belong to. No manual collection, no gap between the two markets, and a working scraper four days after the brief. Over three seasons the client ordered two more full refreshes, the last one with offices included.

See our ImmoScout24 scraper and real estate data scraping, or contact us to plan a full portal export with images.