Lönetracker

A web app for hourly-paid retail workers in Sweden. Upload your schedule, get what the month actually pays — OB (obekväm arbetstid) supplements, tax and net pay, with every hour attributed to the rule that paid for it.

Who it's for

Detaljhandelsavtalet §7.3 settles the pay form: a heltidsanställd is always on månadslön, while a deltidsanställd is on timlön by default, moving to månadslön only if their hours are averaged over 11+ weeks or they agree to it.

Someone on a monthly salary already knows roughly what lands in their account. Someone paid by the hour, on a schedule that shifts week to week with OB spread unevenly across it, does not. That second group is who this is for — part-time store staff, which is most of Swedish retail.

What it does

• Reads a schedule three ways — a .csv/.xlsx export, text pasted from a PDF or email, or shifts typed in by hand
• Splits every shift across OB tiers by day type and time of day
• Handles shifts running past midnight, applying the next day's rules to the hours after 00:00
• Flags shifts over five hours with no break recorded
• Shows gross pay, tax at a configurable rate, and net pay, plus a per-shift breakdown you can expand
• Swedish and English, switchable
• Installable as a PWA; everything is stored in the browser's local storage

Getting a schedule in

Every employer runs a different scheduling system, and a parser can only be written for a format someone has actually seen. So there are three routes in, and the last one always works:

• A file. .csv or .xlsx. Columns are found by their headings, in Swedish or English. If the headings are unfamiliar the file is not rejected — the columns are listed with sample values and you point out which is the date, the start and the end.
• Pasted text. Copy from a PDF, an email or a web roster. Any line with a date and two clock times becomes a shift; a third short time is read as a break. A line that also says sjuk or semester becomes that kind of leave; a semester line needs no times. Dates are parsed strictly, since guessing day-versus-month would move a whole month onto the wrong weekdays, and weekday is what decides OB.
• By hand. Arbete, Sjuk or Semester, then date, start, end, break — a sick day takes the shift that was scheduled, a semester day only the date. Tedious for a month, but it needs nothing from the employer at all — it works for someone whose schedule only exists on paper.

A date is work or leave, never both: adding one removes the other on that date and says so. Any shift in the list can be marked sick or semester, and back.

The OB rules

Taken verbatim from §8.1 of Detaljhandelsavtalet, 1 April 2025 – 31 March 2027 (Svensk Handel / Handelsanställdas förbund):

When	Pays
måndag–fredag kl. 18.15 – 20.00	50 %
måndag–fredag efter kl. 20.00	70 %
lördagar efter kl. 12.00	100 %
söndagar och helgdagar	100 %

Med lördagar jämställs jul-, nyårs- och midsommarafton.

Supplements are paid on top of the base rate: an hour at OB 50% with a base of 177.44 kr/h pays 177.44 + 88.72 kr.

Rules are data, not code

A rule set is a list of tiers (a name and a supplement percentage) and a list of time windows (day types, a start and end time, and the tier they pay). Any hour not covered by a window pays the base rate only. The preset is just the starting point — a different agreement, or a local addition like Coop's extra 05.00–06.00 tier, can be entered by hand.

Day types, not just weekdays

A window targets day types, which is what makes "söndagar och helgdagar" work. Beyond the seven weekdays there are two more:

• Helgdag — Swedish public holidays, computed per year including the Easter-derived ones, midsommardagen and alla helgons dag. A Thursday that is Kristi himmelsfärdsdag pays holiday rates all day.
• Afton — julafton, nyårsafton and midsommarafton, which §8.1 equates with Saturdays. A Sunday stays a Sunday even when it is also an eve, since the Sunday rate is the higher of the two.

Breaks

Detaljhandelsavtalet sets no minimum or maximum length for a rast — the schedule decides. What it does settle is whether the time is paid:

• Rast is unpaid and not counted as working time (§6.1), so it is deducted by default. Because a schedule rarely says when the break fell, it is spread proportionally across the tiers rather than taken off the end — otherwise it would silently eat the best-paid hours.
• Måltidsuppehåll (§6.5, allowed only in exceptional situations) and shorter pauser under arbetstidslagen do count as working time. Tick "Rasten är betald" and nothing is deducted.

Arbetstidslagen §15 entitles you to a rast after at most five hours of work, so any shift longer than that with no break in the file is flagged — usually it means the export left the break out.

Leave

Leave days sit in the export with their scheduled hours intact, so they have to be told apart from work — and from each other, since each is paid differently. A day counts as leave only when the absence columns cover the whole shift; those columns routinely carry a minute or two on ordinary days, meaning someone arrived late.

• Semester — semesterlön is 13 % of the pay for work in the intjänandeår (1 April – 31 March), shared over the paid days, and never below the garantibelopp pro-rated by hours (§14.6). The daily figure on a payslip ("Semesterlön betald ... a") is the one to use, since the employer's underlag includes semesterlönegrundande frånvaro (§14.7). Without one, the panel estimates it from the year's pay and hours and labels the result an estimate. The mid-year starter's two-step pro-rating (§14.6 B 3) is not modelled. Until a daily figure exists the days are counted, paid nothing, and named beside the gross as left out.
• Semesterersättning — only when the person ticks that their payslip shows it every month: 13 % of the month's pay for work (§14.8), not of semesterlön or sjuklön. Otherwise it is paid when the job ends and is not counted.
• Sjukdom — calculated per §15.4. Sick days are grouped into periods, where a fresh spell within five calendar days continues the previous one. Each period opens with a karensperiod whose length in hours is 20 % of the agreed working week, taken off the front and unpaid; the rest pays 80 % of the hourly rate. §15.4 also gives 80 % of the OB those hours would have carried; the one real payslip with sick days on it did not pay that, so the OB part is shown beside the sjuklön and kept out of the gross. A day worked in part and then left sick is both: the hours worked pay as work, the rest of the planned shift pays sjuklön. Days past the fourteenth are Försäkringskassan's and are excluded, with a note. Without weekly hours the karensperiod cannot be sized, so sjuklön is shown as a ceiling and kept out of the gross until they are entered. Not modelled: the cap of ten karensavdrag in twelve months, and the sjukanmälan and läkarintyg conditions (§15.2–15.3).
• Other leave — counted and shown, never included. Tjänstledighet and the rest may not be paid at all and the export does not say which.

Not covered

• Övertid and mertid. §8.1 notes that where overtime and OB overlap you get only the higher percentage, not both. Not modelled as such. Where the export carries a Godkänd column — the hours the employer signed off — that column is what is paid, so approved mertid is counted through it; without the column, hours beyond the schedule look identical to clocking out late and are left out rather than guessed at.
• Månadslön. §8.1 derives the hourly rate as 1/166 of monthly salary. Out of scope by design — see Who it's for.
• Agreements other than Detaljhandelsavtalet. Lager- och e-handel has its own rates and its own structure; rather than ship figures taken from a summary page, there is no preset for it. Build the windows by hand if you need them.

Accuracy

Checked against three real Bestseller time exports and the payslips they produced. June and August 2026 match the payslip to the öre on gross, and to the krona on tax and on what was paid out; July is 1,77 kr under on gross, which is the payslip splitting one month over two stores and rounding each half on its own — something the export does not show. Every OB tier is exact. August is the sick month: two separate sick days, two karens periods, 4,16 h of sjuklön to the öre, and two days of approved mertid read from the Godkänd column.

The app rounds the way payroll does — the month's hours per row to two decimals, the rate to the öre, each row to the öre, the payout to whole kronor — because a krona's difference can cross a skattetabell bracket edge and move the tax by fifty.

Privacy

There is no backend, no account, and no analytics. The schedule is parsed in the browser and the results never leave it.

Sources

• Detaljhandelsavtalet 2025–2027 (PDF) — §6.1, §6.5, §7.3, §8.1, §14.6–14.8, §15.1, §15.4
• Försäkringskassan: karensavdrag
• Handels: OB-tillägg
• Handels: När har jag rätt till rast och paus?

Agreement periods change. If you are reading this after April 2027, re-check the figures against the current avtal.

Tech

Next.js 16 (App Router), React 19, TypeScript, Tailwind CSS 4. Every route is statically prerendered — the app is entirely client-side.

The .xlsx reader is hand-rolled: it walks the zip's central directory and inflates the entries with the browser's built-in DecompressionStream, so the app ships no spreadsheet dependency. Persisted state is exposed to React through useSyncExternalStore, which also picks up changes from other open tabs.

lib/
  rules.ts     tiers, windows, preset, day classification, validation
  holidays.ts  svenska helgdagar, incl. Easter-derived dates
  calc.ts      shift splitting and pay totals
  parse.ts     csv + xlsx readers, schedule column detection
  storage.ts   localStorage load/save
  store.ts     useSyncExternalStore bindings
  i18n.ts      sv/en strings
  time.ts      time parsing and formatting

Running it locally

npm install
npm run dev

Then open http://localhost:3000.

Further reading

• docs/lonebesked.md — reconciling against a payslip, and why pay being a month in arrears trips this up
• docs/skatt.md — skattetabeller, where the data comes from and how to regenerate it for a new tax year
• docs/testing.md — what is proven, what is not, and the browser suite in e2e/