A match profile is the description the software compares candidates against. Build it from the job and it finds people who can do the work. Build it from a person and it finds people who resemble that person.
Start from the scorecard, not the job advert
Your advert was written to attract people. It contains company description, benefits and tone — none of which should affect a ranking.
Feeding the advert in as your match profile is the most common setup mistake. It produces rankings partly driven by how much a CV sounds like marketing copy.
The right input is the scorecard from Module 2 — specifically the items you marked as checkable from a CV. Those are already written as things you could watch someone do, with notes on what counts as proof. That is exactly the shape a match profile needs.
Not all evidence is worth the same
Most tools treat everything in a CV equally. It is not.
| What it says | Worth | Why |
|---|---|---|
| Described work with specifics — scale, numbers, outcome | Highest | Hard to make up, easy to ask about at interview |
| Tools named inside a project description | High | Shows use, not just exposure |
| Job titles | Medium | Every company names them differently |
| A list of skills | Low | Self-declared, unverifiable, easy to stuff |
| Education | Low for most jobs | Weak predictor, and closely tied to background |
| The summary at the top | Lowest | Written to match whatever was advertised |
If your tool lets you weight sections, push weight towards described work and away from skill lists and summaries. If it does not, know that part of your ranking is measuring who wrote the better opening paragraph.
The "just like our best employee" trap
The easy way to build a profile is to take your best team member's CV and use it as the target. It feels sensible. It also builds in everything incidental about that person: the kind of company they came from, how they got into the field, the way they write.
You end up selecting for resemblance rather than ability — including things you never intended and cannot see in the results.
Test for it directly. Ask: if someone had done all the same work at a different kind of company, by a different route, would this profile still find them? If not, rewrite it around what was done rather than where.
Two tests, ten minutes
One strong, one weak. Pick two people whose suitability you are sure about — ideally past applicants for a similar job. Run both. If the strong one does not rank clearly above the weak one, your profile is not measuring what you think.
The shape test. Take your top 20 and look at their careers as shapes rather than contents. Are they all continuously employed? All from one industry? All from big companies? All following the same route in?
A profile that only finds one shape of career is selecting on the shape — and career shape reflects circumstances far more than ability.
Expect to rewrite it
Two or three attempts is normal. The usual corrections, in order of how often they come up: remove a requirement that was really a preference, swap a job-title rule for a what-they-did rule, cut the weight on skill lists, and add something that separates your strong example from your weak one when the first run put them close together.
Keep the profile with the job
Store it alongside your screening rules and the shortlist. It is part of the record of how a decision was made, and it is what you will want when you hire this job again in eight months.