Classifieds data is harder than retail data, and the reason is that there is no catalog.
A retailer has SKUs. Every listing maps to a known product with a known description. A classifieds site has whatever a stranger typed into a form. The same car appears as “VW Golf 7 2.0 TDI”, “golf mk7 diesel” and “Volkswagen Golf, 2019, excellent condition”. There is no identifier tying them together, and half the useful information is in free text.
This guide covers how the Njuškalo Croatia API and the Gumtree UK API model that problem.
Category trees do the work
On a classifieds site the category is the schema. A listing in “Cars” has mileage, year, fuel type and transmission. A listing in “Property” has rooms and floor area. A listing in “Furniture” has neither. Attributes are defined per category, so you cannot write one parser and be done.
Both APIs expose this explicitly, which is what makes them workable.
Njuškalo gives you /category-navigation for the tree, /category-detail for one node, and /filter-options for the selectable values on a category’s attribute filters. Plus /recommended-categories and /trending-categories, which tell you where activity is concentrated right now.
Gumtree gives you /categories for the tree or a subtree, /filters for the filters available on a category, and /filter-options for the values of a single attribute filter.
The workflow is the same on both: walk the tree, pull the filter definitions for the categories you care about, then build your extraction per category. Trying to search first and infer structure later is the mistake.
Search, and knowing what you missed
Njuškalo has a genuinely useful pair. /search returns results; /search-count returns just the count.
That second one solves the coverage problem. Classifieds search endpoints cap results, so “every car listing in Croatia” is not one query. Calling /search-count first tells you whether your filter combination is under the cap, and if not, you split it by price band or region and recurse. Without a count endpoint you are guessing, and you find out you were wrong when your dataset is quietly 60% complete.
/search-by-category scopes to a branch, /latest returns the newest listings, and /parse-url turns a Njuškalo URL into structured query parameters - useful when someone hands you a link rather than a spec.
Gumtree covers /search and /search-by-url, plus two demand-side endpoints: /search-suggestions for keyword autocomplete with matching categories, and /trending-searches for what people are currently looking for. Trending searches are a demand signal you cannot derive from listings, which are supply. Together they tell you where demand is outrunning inventory, which is the interesting part.
Sellers, and why they matter
Both APIs expose seller data, and on a classifieds marketplace this is often where the commercial value sits.
Njuškalo has /seller-profile and /seller-reviews. Gumtree has /seller, /seller-listings and /seller-reviews.
The reason to care: classifieds markets contain a mix of private individuals and professional dealers operating as if they were private. Listing counts per seller, review history and posting cadence separate the two. For anyone doing market analysis this is essential, because dealer pricing and private pricing follow different logic and mixing them produces a bimodal mess that averages into nonsense.
It is also the basis of lead generation, which is a common use of both platforms.
Geography
Njuškalo has /geocode and /reverse-geocode. Gumtree has /locations-autocomplete, /locations-nearest for reverse geocoding a coordinate, and /countries.
Both let you go from a coordinate to a location the platform recognises, which is what map-based search products need. Resolve to platform location IDs rather than matching place names.
The two markets
Croatia (Njuškalo) is the dominant marketplace in a small country, which makes it unusually complete. When one platform holds most of a national market, its listings approximate the market rather than sampling it. Cars and real estate are the deepest categories. Croatian property has a strong coastal and tourism dimension, so seasonality is real. /popular-brands is a useful shortcut into automotive analysis.
United Kingdom (Gumtree) is a large, competitive market where Gumtree sits alongside eBay, Facebook Marketplace and category specialists. It is a sample of UK second-hand activity rather than the whole of it, and it is strongest in cars, home and garden, and general goods. /similar-listings supports recommendation features, and /listing-by-url handles the case where a user pastes a link.
What people build
Price monitoring for second-hand goods and vehicles, which requires separating dealer from private listings first.
Market intelligence. Supply by category over time, cross-referenced against Gumtree’s trending searches for the demand side.
Lead generation. Sellers posting repeatedly in a category are businesses, and identifiable as such from listing counts and review history.
Aggregators and search products that pull several sources into one interface.
Demand research. Trending searches and category volume as a read on consumer intent, which moves faster than retail sales data.
Recommendation engines built on similar-listing endpoints.
Practical notes
Normalise before you compare. Free-text titles need cleaning, and category attributes are more reliable than anything parsed out of a description.
Check for a count endpoint before assuming your extraction is complete. If there is one, use it.
Listings are removed as well as added. If you only ever add, your dataset will drift toward a market that no longer exists. Track disappearance as its own event, since it is a weak signal that something sold.
Prices are asking prices, and on classifieds the negotiation gap is wider than in retail.
Related APIs
For another multi-vertical classifieds source with registered transaction prices, see the Yad2 Israel API. For UK property specifically, the Rightmove API is deeper than Gumtree’s property category. Browse the full API Library.
Disclaimer
Happy Endpoint is not affiliated with, endorsed by, or sponsored by Njuškalo or Gumtree. All data is collected from publicly available sources.
