Transparency

What is automated

• Data collection: 27 datasets from 10 public-sector providers are synced automatically on schedule — Skolverket daily, SCB and Kolada monthly. Live sync status per data source is on the settings page.
• Calculations: Merit scores, SALSA scores, eligibility rates and all other key metrics are calculated deterministically from source data. Calculation logic is documented on the method page.
• Page generation: All pages are generated automatically at every deploy (Astro SSG). No pages are manually edited after generation.
• Archiving: Data is automatically versioned with timestamps at every update. See the versioning policy.
• Quality control: Sync guards compare the record count in new data with the previous version and block publication if the count drops by more than 20%.

What is manual

• Editorial content: Interpretation guides, glossary, data stories and this page — written by Markus Reimer.
• Reasonableness checks: Automatic checks are supplemented with manual review during large data changes.
• Corrections: Reported errors are investigated and fixed manually. See the corrections policy.
• Preschool geodata: Preschools without coordinates in Skolverket's register are matched against SCB's geodata and OpenStreetMap's Nominatim. About 85% have been matched.
• School Inspectorate data: Supervisory results, injunctions and fines are synced quarterly on schedule, plus via manual trigger when needed.

Potential biases and limitations

• No selection: All open school units in Sweden are included (dormant and discontinued excluded) — there is no selection that could create systematic bias.
• Timing: Data may be 1–12 months old depending on the source. Grade data refers to the previous school year and is published with a delay.
• Statistical confidentiality: Skolverket's confidentiality rule (<15 pupils) hides data for small schools. This means small schools systematically lack more data points.
• Methodology choices: Choice of SALSA model, weighting in the school choice guide and aggregation methods are documented but involve unavoidable design decisions. See the SALSA method and method policy.
• School survey: Response rates vary widely between schools. Results with fewer than 5 responses are not shown; remaining values should be read together with response rate and respondent count.

Independence and conflicts of interest

Skolkoll is operated by Skolspegeln AB (company reg. no. 559359-7288). For transparency, it is declared that Markus Reimer works professionally at AcadeMedia, Sweden's largest independent education provider.

Skolkoll is an independent project with no resourcing, editorial control or influence from that employer. All data comes from open public sources and is processed identically for all schools and providers.

If a school provider wants to "look better" on Skolkoll, there is only one recipe: deliver good quality that is reflected in the public statistics.

Open data

• All data compiled by Skolkoll is open and free to use. The codebase is not currently published as open source, but calculation methods and data sources are documented openly on the method page.
• Skolkoll's compilations are published under CC BY 4.0.
• Source data follows the respective data source's licence (Skolverket, SCB, Kolada etc.).

AI on Skolkoll

AI is used in a few, well-defined places — chiefly the assistant Kollen, but also to stylise school images and as internal drafting aids. Two things hold everywhere: the AI sets no published numbers(all statistics are computed deterministically from source data) and nothing AI creates is published or sent on without human approval (the Kollen chat aside — a clearly labelled AI assistant you choose to talk to). SeeHow Skolkoll uses AI for a full account of where AI is used, what controls are in place and what the AI does not do.

← About Skolkoll · Method · How Skolkoll uses AI · Corrections policy · Roadmap