# Salary transparency in Swiss job ads: what a missing field hides

A structured salary field was populated in just three records in Panta’s frozen Swiss-location cohort. Reviewing stored source text and metadata confirmed numerical remuneration in at least 230 canonical records, including training stipends, total compensation and a contractor task fee. This exposes a measurement gap, rather than establishing a national salary-disclosure rate.

By Panta. Published 2026-11-05.
Canonical: https://withpanta.com/insights/salary-transparency-switzerland/

## A salary field can miss pay in the advert

230 reviewed records with confirmed numeric compensation in the screened cohort. The article separates role pay, total compensation and task fees. This is a metadata audit at 4 November 2026, not a Swiss disclosure rate.

| Evidence category | Listings |
| --- | ---: |
| With a structured salary field | 3 |
| Without a structured salary field | 227 |

## Can you see the pay before applying?

Sometimes the salary is already in the advert, even when a job search database appears to have no salary information. That distinction matters to applicants comparing jobs and to anyone publishing a salary-transparency statistic.

In Panta's frozen **4 November 2026** cohort of **14,536 canonical records classified as physical Swiss workplaces**, only **three** contained a populated structured salary field. A numerical-pay screen followed by review of the stored evidence identified **at least 230 records with numerical remuneration for the advertised work or programme**. **227 of those records had an empty structured salary field.** The other three carried pay in source-derived structured metadata.

These are records in Panta's indexed inventory, including recruitment-agency adverts, training programmes and staffing pools. They are not 230 independently verified vacancies or an estimate of how many Swiss employers disclose pay. The larger lesson is that **missing salary metadata cannot establish that an employer withheld the salary**.

## What the review counted

We screened title, stored description and salary metadata using explicit currencies, plausible numerical amounts, pay vocabulary and rate units. Every one of the **415 screened records** received an evidence classification. We also inspected pay-related and numerical passages in **48 deterministically selected records outside that screen**, split between those with other currency mentions and those without. This additional check found no further numerical role pay; it does not establish complete recall.

| Evidence confirmed in the screened records | Records |
| --- | ---: |
| Numerical role pay, including training remuneration and stipends | 227 |
| Explicitly labelled total compensation | 2 |
| Contractor fee per accepted task | 1 |
| Total included in the numerical-remuneration finding | 230 |

The 230 records span **43 recorded company or recruiter identities**. These identifiers are not a count of independent employers: one organisation can have more than one recorded identity, and a recruiter can advertise for several clients. Canonical records can also repeat the same real opportunity. We retain that uncertainty rather than silently deduplicating by an invented rule.

The two largest recorded groups were **Novartis, with 35 records**, and **ZFV, with 31**. Removing the largest group still leaves **195 confirmed records**. This supports the existence of a broad extraction gap across several sources; it does not make the corpus representative of Switzerland.

## A salary amount still needs a denominator

The evidence includes very different things that a single “salary available” checkbox would conceal:

- **Full-time equivalent pay.** ZFV adverts state annual compensation at a 100% workload, including the thirteenth salary. A part-time title does not mean the full annual figure is the part-time offer. PostFinance's captured adverts also identify the range at 100% workload.
- **Hourly role pay.** The stored USZ nursing-pool adverts state hourly rates. Mimic Robotics' robot-operator and data-collection adverts do too. An hourly amount says nothing by itself about guaranteed hours or annual earnings.
- **Training and programme pay.** Apprenticeship schedules specify different amounts by training year. CERN's graduate and fellowship adverts use net stipends or allowances with their own conditions. They are not directly comparable with ordinary annual gross salaries.
- **Base pay and total compensation.** Jua's source metadata labels its range as total compensation and mentions an equity plan. We preserve that label rather than treating the range as guaranteed cash base salary.
- **Contractor output fees.** One Lightly advert offers a USD fee per accepted task. The advert's estimated completion time cannot be converted into a guaranteed hourly employment wage.

We therefore publish no median “Swiss advertised salary” from this collection. Annualising all these amounts would require assumptions about hours, workload, payment frequency, variable compensation, taxes and programme status that the stored evidence does not consistently resolve.

## What we kept out

A currency mention is a useful lead, not a salary finding. The screen surfaced **185 records outside the included remuneration categories**:

| Reason for separation | Records |
| --- | ---: |
| Benefit, premium, business amount or other non-base-pay number | 144 |
| No numerical role pay detected after review | 30 |
| Salary explicitly attributed to a foreign location | 4 |
| Indicative recruiter remuneration or commission-led earnings possibilities | 3 |
| Unusable zero-to-zero salary placeholder only | 3 |
| Currency could not be resolved | 1 |

The categories sum to **185 excluded or unresolved records**, alongside the 230 included records, reconciling all 415 screened records. These counts describe our screening process, not a ranking of employers.

For example, a laptop voucher, discounted lunch, corporate revenue figure or night-shift premium does not supply a role's base salary. US salary text attached to a multi-location corporate description cannot establish its Swiss pay. A posting that combines a euro symbol with “CHF” remains ambiguous. Several zero-to-zero salary headers were unusable, while other zero headers were contradicted by useful apprenticeship-pay schedules further down the same description. We counted the latter from the actual schedule, not from the placeholder.

## Why published transparency figures can disagree

Other companies already publish useful material about salary visibility. Their definitions need to travel with their numbers.

[jobs.ch reported in October 2025](https://www.jobs.ch/de/job-coach/ueber-geld-spricht-man/) that around **40%** of adverts on its platform displayed salary information in August 2025, compared with under 2% in January 2023. That is a platform observation at different dates. [JobCloud's terms, updated in June 2026](https://service.jobcloud.ch/hc/en-us/articles/45031646738321-General-Terms-and-Conditions-GTC), also allow it to add clearly identified market-based salary information where the advertiser supplies none. Those terms do not establish how the earlier 40% was composed. They show why displayed pay and advertiser-supplied pay should be measured separately. Our audit cannot be compared with that percentage as a provider-performance score or a trend.

[SwissDevJobs asks advertisers to provide salary ranges](https://swissdevjobs.ch/frequently-asked-questions). Its technology listings are useful for researching explicitly advertised ranges, but a portal designed around salary transparency is a selected population, not a neutral sample of every Swiss job advert.

[PostFinance announced salary ranges in its adverts from 2024](https://www.postfinance.ch/en/blog/people-purpose/pay-transparency.html), with senior management excluded from the policy described in its interview. Our two captured PostFinance records contain ranges in the body. They illustrate a published practice; two observations do not audit compliance across the organisation.

An [October 2026 IZA discussion paper](https://www.iza.org/publications/dp/18992/job-seekers-beliefs-and-preferences-over-wages-and-benefits-experimental-evidence-from-online-job-boards) takes a different approach: researchers randomly displayed wage and benefit information from an employer-review platform to 265,000 users of a Swiss job-board network. It studies how information changes job-ad engagement and beliefs. Panta's descriptive audit observes stored advert evidence; it does not estimate an effect on applications, hiring or pay equality.

## What this means for your search

Open the original advert and check whether the figure refers to annual or monthly pay, actual or 100% workload, base salary or total compensation, and fixed pay or an earnings possibility. If the page shows a platform estimate, keep it separate from a range supplied for that position. If the information is absent or unclear, a useful early question is: “What is the gross base-salary range at my intended workload, and which payments or variable components sit outside it?”

For Panta's research, the next step is a source-aware salary field that retains evidence, units and uncertainty, followed by a genuinely independent validation sample. Repeated snapshots could then test change within a stable source panel. This first audit establishes neither a national disclosure rate nor a historical trend.

## Method and limits

The input is the same immutable snapshot used by our [November baseline](/insights/swiss-job-market-2026-11/), restricted to physical-CH records with a fresh source observation. **871 records lacked stored descriptions.** Even a nonempty description can omit content present on the original page. External salary scales, linked policies and interactive salary calculators were not followed for every listing.

The screen was revised during evidence review to recognise Swiss-franc wording, compound German pay terms and small hourly amounts. This is iterative AI-assisted research, not a held-out accuracy benchmark or independent human double coding. The public pack retains classifications, source links, hashes and evidence offsets without republishing descriptions or personal contact details. The [methodology](/insights/methodology/) explains the wider corpus limitations. Our [80% careers report](/insights/eighty-percent-careers-switzerland/) separately examines the workload assumptions that salary comparisons often overlook.