Unpaid work on a résumé, and the twenty points our checker used to take off for one word

Care and voluntary work is real work described in a vocabulary that documents and parsers both handle badly. We measured exactly how badly, found the fault in our own rule, and measured again after fixing it.

ConxfolioPublished 2026-09-12 · updated 2026-09-13

— the role first. the status in brackets after it. it is better writing either way.

Care work, voluntary work and unpaid caring are the largest category of real work that never makes it onto a résumé properly. Part of that is a vocabulary problem: the roles have titles nobody outside the sector recognises. Part of it turns out to be mechanical, and we were able to measure that part exactly. The résumé builder and the checker were the instruments.

The page we used

An invented final-year student with no paid employment: a dissertation, a mapping project, a volunteer stewarding role and a student representative role, one degree, seven skills. 169 words. Both unpaid roles sat under an EXPERIENCE heading, with dates and an achievement line each. That is a normal, honest page and it is how most people write unpaid work down.

The first result, on 12 September

The ATS résumé checker read 169 words, 19 lines and five recognised headings. Parseability scored 100. Completeness scored 70, and the failed check said: no experience section with role, dates or achievement evidence was detected. Content scored 12, because the three rules that read experience lines had no experience lines to read. The weighted total was 56 and the published score was 55, because a ceiling applied.

The cause, which was one word

The first line under the heading began with Volunteer. The checker's section rule was treating a line opening with that word as the start of a new section, so everything beneath it was filed elsewhere and the experience section the scorer looked at was empty. The heading was there. The roles were there. The rule stopped reading at the word.

The second result, on the same day

Route Steward (volunteer) instead of Volunteer route steward. Completeness went from 70 to 100, content from 12 to 37, the ceiling lifted, and the published score went from 55 to 75. Same 169 words, same 19 lines, same five recognised headings. We also tested moving the whole experience block above projects, which changed nothing at all, so the word was the variable.

The rule was changed, and the page re-measured

A twenty-point difference between two ways of writing the same job was a defect, so the heading rule was rewritten the next day. It now asks whether a line is structurally a label: no dates in it, no entry separators between its own words, no verb clause after the label, and a shape that reads as a label on its own line. Volunteer route steward, Manchester Cycling Campaign, 2024 to 2025 carries dates and separators, so it stays where it was written.

Re-measured on 13 September, the original page scores 100 × 30% + 100 × 30% + 62 × 40% = 85, and so does the rewritten one, and so does the reordered one. The 55 and the 75 above are kept because they are what this page measured when it was written, and because the reason a rule changed is worth more than the number it produced.

What to take from that

Write the role first and the status in brackets after it. Support Worker (bank). Ward Volunteer becomes Ward Assistant (volunteer). Carer for a family member is written as a period with a duration rather than as a job title. This is not gaming a score: it is the ordinary rule that a job line starts with the job, and it happens to be what parsers read. The longer version is in a student résumé with no paid jobs on it.

The evidence care work has

Hours covered, people supported, a rota you ran, training you completed, a process you changed. The measurable rule reads percentages, currency, multipliers and a list of unit nouns. Patients, residents, clients, members and shifts were absent from that list on 12 September and were added on 13 September, which is the same repair a nursing page prompted, set out in a nursing résumé full of numbers. Never invent a figure to satisfy a rule.

Where the caring period itself goes

The product has no field for it and takes no position on whether it belongs on the page. Where you want it there, the additional section takes a heading you name. Where you want it explained rather than listed, the cover letter builder has an optional field for what only you know, in your own words, and it will not invent a sentence for you. How to explain a career gap covers both choices.

What we are not claiming

That a higher number changes an outcome. The checker is not an employer's applicant tracking system, it uses no job description and it predicts nothing, and no page here says a document gets anybody hired. What is claimed is two measured facts: on 12 September that word cost twenty points on that text, and on 13 September, after the rule was changed, it costs none. The rewrite is better writing regardless, and other people's parsers are not ours.

One more thing about how the review behaves on a page like this. The import review prints what the reading found and reports the fields it could not fill rather than filling them, so a page of unpaid roles arrives in the editor with its gaps visible. Target role, certifications and languages all came back as not found on our test page, which is the correct answer and a more useful one than an invented guess.

That is the whole design position, and it is the reason a number like 55 is worth reading rather than resenting. The product tells you what it could not see, in the report's own words, which is how the misreading above was found at all. Where what it could not see was real work, the fix is sometimes in the wording and sometimes, as here, in our rule. Where what it could not see was genuinely absent, no rewording will help and no honest tool should pretend otherwise.