Coding agent documentation index Fetch the complete documentation index at: https://docs.worklittle.com/docs-agent-manifest.json Use this file to discover all available pages before exploring further.

How fit scores work

What the % after a candidate's or employee's name means, what goes into it, and how to use it fairly.

The 67% after someone's name on Candidates and Employees is a fit score: a quick read on how well a person's background lines up with a job. It's worked out by a state-of-the-art AI model plus a few simple checks, and it's meant as a starting point, never the only thing you go on.

Use it as one signal, not a decision

A fit score helps you see relevant experience, skills match, and the other factors below at a glance, so you know where to look first. It can be wrong: résumés leave things out, people describe the same work in different words, and the model only sees what's written down.

- Read the application or profile before you decide anything. Never reject, hire, promote, or let someone go because of the number alone. - Worklittle never rejects or hides anyone because of it. It's one of several signals in the Relevance sort on Candidates, and you can sort any other way you like. - No number doesn't mean a bad fit. It means there wasn't enough to judge fairly, so review that person yourself.

What goes into the score

The score is a weighted average of these parts. Hover the % (or tap it on a phone) to see each part's own score.

| Part | Weight | What it looks at | Scored by |
| --- | --- | --- | --- |
| **Relevant experience** | 40 | How closely past work matches the job's actual duties and level, counting only relevant years toward any years the post asks for | AI model |
| **Skills match** | 25 | How many of the skills, tools, and knowledge the job needs the person shows | AI model |
| **Screening questions** | 15 | Answers to questions you gave an ideal answer, checked against that answer | Simple check, no AI |
| **Education** | 10 | Whether degree level and field meet what the post asks for. Only used when the post mentions education | AI model |
| **Application answers** | 10 | Whether written answers (like a cover letter) address the question and show they can do the job. Substance counts, not grammar or polish | AI model |
| **Performance feedback** | 30 | Employees only: what written performance feedback on their profile says about how they do this role's work | AI model |

When a part has nothing to go on (no screening questions, no written answers, no feedback), it's left out and the rest are scaled up, so a missing part never counts as zero.

Which job it's scored against

- Candidates: the job they applied for. Someone who applied to several jobs shows their strongest fit on a job that's still open (or their best ever if none are open). Filter to one job to see that job's score. - Employees: the role they hold now, or for Alumni the role they had. That's the job post they were hired from, else your job post with the same title, else just their job title. Hired people on Candidates show this score when they have no scored application.

What it never uses

Only job-related qualifications reach the AI model. Before anything is scored, Worklittle removes:

- Names, email addresses, phone numbers, links, and photos - Birthdays, ages, and four-digit years (graduation years can reveal age) - Street addresses and where someone lives - School names (only degree level and field count, not prestige) - Answers to demographic questions

The model is also told never to guess at gender, race, ethnicity, nationality, religion, disability, pregnancy or family status, or appearance, and not to mark people down for gaps between jobs, part-time work, career changes, or non-native English. When something isn't mentioned, it scores that part in the middle instead of assuming the worst.

When it updates

- Candidates are scored right after they apply. Scoring… means it's still being worked out, usually within a minute. - Employees are scored within about 30 minutes, and again after you change their title, skills, or education, or add performance feedback. The old number shows until the new one is ready.