FRED · as of Aug 14
The Bag™

How the numbers work

Every score in this app is a weighted blend of seven things, each of which is shown to you separately. There is no hidden aggregate and no model deciding privately how it feels about you — the arithmetic is below, and you can do it yourself.

The seven components

  • Direct capability fit

    weight 0.30

    How much of what the job asks for you have already demonstrably done. Matched against the capabilities extracted from your own history, not against your job titles.

  • Transferable fit

    weight 0.15

    The parts you have not done under that name but have done under another. This is the component that makes a career change legible, and it is judged with evidence rather than assumed from adjacency.

  • Compensation fit

    weight 0.15

    What it pays against what you need. Where the employer does not state pay, this is unknown — not zero, not average.

  • Seniority fit

    weight 0.10

    Whether the level matches where you actually are. Guards both directions: a role far beneath you is a bad match too.

  • Location fit

    weight 0.10

    Where the work physically is, against where you can physically be — including whether a listing is genuinely remote or merely says so.

  • Credential readiness

    weight 0.10

    Licences, certifications and registrations the role requires, and how far you are from holding them. Knows the difference between never held, in progress, and lapsed.

  • Evidence strength

    weight 0.10

    How well you could actually prove the above to somebody else. A capability you cannot evidence is a capability you cannot get paid for yet.

Weights sum to 1.00. Scorer v1.0-deterministic — the version is stored on every score, so an old number is never quietly compared against a new rule.

The most important thing on this page

A dash is not a zero. Unknown is not bad.

When the app cannot judge a component — the employer did not state pay, the posting never named a location, nothing in your history speaks to the requirement — it shows — rather than a number, and removes that component from the blend entirely, redistributing its weight across the components it can actually judge.

This matters more than it sounds. The obvious implementation scores an unknown as a neutral 50, and then a 50 meaning “we have no idea” looks exactly like a 50 meaning “you are a mediocre match” — same number, same bar, opposite meanings. Silence from an employer is not evidence against you.

What the bands mean

  • 95+ExceptionalRare. Drop what you are doing.
  • 90+StrongA real match on the parts that matter.
  • 80+Worth exploringGood reasons to look, and something to close.
  • 70+StretchReachable, but you would be arguing for yourself.
  • below 70Below the lineShown so you can see it was considered, not hidden.

Scores at “worth exploring” and above breathe — the slow pulse on a numeral marks the ones worth your attention. It is styling, not a separate signal.

How a match is actually made

  1. 1

    Your history becomes capabilities, not titles

    What you upload or say is broken into things you can do, each anchored to the specific fact that demonstrates it. "Product Manager" is not a capability; "ran PI planning across four teams" is.

  2. 2

    Postings are classified into lanes

    Every opening is sorted into a role class — a kind of work, not a job title — so that 3,000 differently-worded postings for the same job stop looking like 3,000 different careers. Lanes are a fixed taxonomy that only grows through a review gate.

  3. 3

    Requirements are extracted from the employer's own words

    What the job actually demands, quoted rather than paraphrased, so a flag or a gap can be checked against the posting and overruled if the reading was wrong.

  4. 4

    The seven components are scored independently

    Each one produces a number and a sentence explaining that number. The sentence is generated from the same facts as the score, so they cannot disagree.

  5. 5

    The blend runs over what is known

    Unknown components drop out and their weight is redistributed. The result is banded, stamped with the scorer version, and stored.

  6. 6

    Nothing reaches a page unattributed

    Validators reject any claim that cannot be traced to a fact id. This is why the app sometimes says less than it could — it will not say more than it can stand behind.

Getting the most out of it

  • Feed it the unglamorous work

    Side projects, volunteering, the thing you built for a friend, the team you ran that was not in your job description. Capability comes from what you did, and the app cannot score what it has never been told.

  • Say what you actually need

    Your floor on money and where you can be are inputs, not preferences. Leaving them blank does not open up the map — it turns two components off.

  • Argue with it

    Every number has its reasoning next to it. Where the reasoning is wrong the number is wrong, and that is a bug worth reporting rather than a verdict worth accepting.

  • Read the components, not just the score

    A 78 that is strong on capability and unknown on pay is a completely different opportunity from a 78 that is mediocre at everything. The headline number cannot tell you which you are looking at.

Still not clear on something? Ask directly — how a number was reached is a fair question and there is a real answer to it.