Where exactly do the deadlines come from?
From the European Commission's public Erasmus+ App API, where each institution reports its own calendar. The entire census —6,622 institutions— is harvested and normalised without altering any dates. What you see on a page is what that university reported, not our interpretation.
Why are only 1,755 universities published if the census contains 6,622?
Because only 1,755 (27%) report any deadline. A page without a deadline would have nothing to say, so it is not published: we prefer a smaller catalogue where every entry is useful, rather than a bloated one full of blank pages. Just because a university isn't listed doesn't mean it doesn't host Erasmus students.
How does "15/05" turn into an actual date?
Institutions report day and month without a year, because deadlines repeat each academic cycle. The site calculates the next upcoming occurrence of that combination from today, and recalculates daily. That is why a countdown is never off by more than a day, and why our exports provide both fields separately: the raw reported day/month, and the calculated upcoming date.
Why do some cities lack living cost data?
Because the figure comes from student responses in the Erasmus+ App, and is only published once at least three responses exist. The catalogue median is 1 per city: with a single response, the "average" would just be one person's budget presented as an entire city's cost. Of 1,250 census cities, 242 meet the threshold. For the rest, the site clearly states that data is unavailable, rather than guessing.
What is the deal with universities where all four dates are identical?
There are 51 institutions where all four procedures fall on the exact same date. It might look like a form error, but it isn't: upon inspection, most are music conservatories and art academies with a single annual audition or call. The page detects this and explains it rather than repeating four identical rows.
Has any data been corrected manually?
Only geographical coordinates: 5 records came with coordinates placing them on another continent and were corrected, with each change tracked in code. Not a single date, email address, or amount has ever been altered by hand.
Do you use AI to generate the page content?
Not to produce data. Deadlines, contacts, and figures come directly from the harvested catalogue, and each page's text is composed from those values. The project rule is that a model can structure existing text, but never search for or invent data: tests showed that asking it to search catalogues produced plausible-sounding fake courses, and an invented course name sent to a coordinator can cost a student their place.