Methodology

Pulse counts new postings across the major Swiss job boards JobHiker tracks, plus a capped sample from LinkedIn. It is not a census of every vacancy in Switzerland — structure and trend, not a full count. We deliberately don’t name individual boards here or anywhere else in Pulse; what matters is the shape of the market, not which specific site a posting came from.

A posting belongs to the Europe/Zurich calendar date it was posted on, or the date it was first crawled when the board gives no posting date. A posting only counts on the day it was crawled within 1 dayof going up — one first seen well after it was posted (a backfill, a keyword search reaching back) does not land on its original day and inflate it after the fact. Excluded rows are counted separately, never silently dropped.

One posting is one row: the same vacancy cross-posted to more than one board is collapsed to a single count at the point it's collected, before Pulse ever reads it.

Every number is compared against the average of the previous 10working days — Monday to Friday, public holidays skipped — and index 100 is that average. Weekends are not in the baseline and are not compared, because a Saturday set against working days would always read as a collapse. So the Monday edition, which covers Friday, Saturday and Sunday, counts all three days in its total but compares Friday only. A comparison only publishes when all of the baseline days have data, their average is at least 20postings, the day itself is not a public holiday, and no crawl change (see below) falls inside the baseline. A holiday inside the baseline is simply left out of it. When a day doesn’t clear that bar, Pulse says so rather than publishing a number that would mislead.

The headline of a day is the biggest change that is real, not just the biggest. A department or canton is reported only when it moved by at least 10% against its baseline, the move is more than 3times the baseline’s own day-to-day variation (so a small department that swings by half every other day does not make the news), and no single employer explains most of the rise — one company listing hundreds of jobs in a day is a batch, not a market move. On an ordinary day nothing clears all three, and the edition says so instead of inventing a headline.

A day's first count, published the next morning, is a lower bound: some postings are still arriving. Revised counts are taken later (typically one and three days after) and stored alongside the original — the original published number is never edited, but the archive's later view of a day can differ from what first went out.

Swiss national and cantonal public holidays are flagged. A holiday is never compared on the day itself, so a quiet holiday Monday doesn't read as a market shift, and a holiday inside the baseline is left out of it rather than pulling the average down.

A crawl break is a day intake changed for a reason on our side — a new source, a scraper fix, a taxonomy change — not a market reason. Trend comparisons refuse to span one. A day marked “single day” is different: something odd happened on that day only, so it is not compared and is left out of every baseline, and the days around it are unaffected.

• 30 September 2026 (single day) — A single employer listed several hundred postings on this day, which is not a market move; this day is not compared and is left out of every baseline.
• 16 September 2026 — PR #52 merged: ISCO-08 taxonomy with real search terms and continuous keyword lanes changed crawl intake.
• 30 August 2026 — Keyword sweep started reaching JobCloud free listings: intake stepped from ~80 to ~7,000 new jobs a day for crawl reasons, not market reasons.

Pulse never publishes individual job rows, full job descriptions or full company lists — only counts, shares and an index derived from them.

If something here looks wrong, or you have a question about how a number was derived, email hello@jobhiker.com.