The XML validator checks a feed against its XSD schema — syntax, structure, content model and data types — for the XML files that transmit data to Yandex services.
The limitation, stated plainly
It does not check the set of elements in the feed. A file missing an element that a partner programme requires will pass validation and then fail to work in the programme. Yandex says outright that you must check the elements manually against the programme’s technical requirements.
This is the trap: a green result here means well-formed, not fit for purpose. Teams treat validation as sign-off and then spend weeks wondering why nothing appears.
The schemas available
- Doctors — a YML feed of doctors and medical services (the default selection).
- Realty — the export format for Yandex Realty listings.
- Videos — video data for Yandex Video.
- Market — the YML export for online store search.
- Business — branch data for bulk addition to Yandex Maps and search snippets.
Selecting the wrong schema produces an error, and because Doctors is the default, the first validation attempt of a Market feed usually fails for that reason alone.
How to use it in practice
- Select the right schema first.
- Validate for structural correctness.
- Then check element coverage by hand against the programme’s requirements — the step the tool does not perform.
- Re-validate after any generator change, since feeds are produced by code that changes for unrelated reasons.
Where it fits
Feeds are how large, frequently changing datasets reach Yandex, as opposed to markup for small stable data. That makes them infrastructure: generated automatically, rarely inspected, and prone to drifting out of specification without anyone noticing.
How we apply this
The element-coverage gap is what we plan around: validation proves the file parses, not that the programme can use it, so we keep a checklist of required elements per programme and check it separately. We also re-validate feeds after any release that touched the generator, because feeds break for reasons unrelated to search work and nobody looks at them until the listings quietly disappear.
Related services