This is the step where a job description becomes something software can act on. It is also where teams quietly build in a requirement they would never defend out loud.
You get three must-haves
A must-have removes people. Three is the limit.
The test for each one: would I genuinely reject an outstanding candidate for missing this? Not "is it important" — everything on the list is important. Reject. An actual, final no.
Most lists fail immediately. Go through yours and you will usually find one or two real must-haves, a few strong preferences, and two or three items that are only there because the last version of the advert had them.
If your list has more than three, you do not have requirements. You have a description of a person who does not exist, and you are about to spend six weeks finding that out.
Everything else becomes a score
A score moves someone up or down the list. It never removes them. Each one needs two things:
- A weight — how much it moves them compared with the others.
- A note saying what counts as proof. "Has shipped a product to real users" can be checked. "Passionate about quality" cannot, and will end up being scored on writing style.
Setting the weights
Share out 100 points between your scores. Forcing a total makes the trade-offs real, instead of everything being "high priority".
Then check three things:
- Is anything above 30? Then it is nearly a must-have. Decide which it is and move it.
- Is anything below 5? Remove it. It cannot change an outcome and it makes the whole thing harder to reason about.
- Do your top three add up to more than 60? Good. If everything is roughly equal, your rules will not separate anyone — you will see that in Module 4 as a flat set of scores.
Test it on two real people
This check catches most mistakes and takes ten minutes.
Pick two people you hired into a similar job: one who worked out well, one who did not.
Score both against your must-haves and weights, using only what was on their CV when they applied — not what you know about them now.
| What happens | What it means | What to do |
|---|---|---|
| Good hire passes, poor hire does not | Your rules are picking up something real | Carry on |
| Both pass easily | Your rules are not separating anyone | Add something that tells them apart |
| The good hire fails a must-have | That must-have is wrong | Turn it into a score |
| Neither passes | Your rules describe nobody you have ever hired | Go back to the intake meeting |
The third row is the common one, and the most valuable thing this exercise produces. A must-have that your best hire would have failed is not a requirement. It is a habit.
When the result disagrees with you
Sometimes the scoring puts someone high that you would not have picked, or puts your favourite low. That is information, and there are only two honest explanations: your rules are wrong, or your instinct is. Both happen.
What is not honest is adjusting the weights until your preferred candidate wins, then presenting the result as objective. If you want to overrule it, overrule it openly and write down that you did. A recorded override is defensible. A quietly rewritten weighting is not.
Download the weighting worksheet · all templates
The line to hold
Must-haves remove people. Scores put them in order. Anything you cannot prove from a document belongs in neither — it is an interview question. Putting it in your screening means scoring people on how well they write about themselves.