There are two ways to screen a pile of applications, and they are not equally fair. The first is to rank people against each other and take the top of the stack. The second is to read each application against the role and say, on its own terms, how well the evidence matches what the job needs. They sound similar. They are not.
Ranking against each other imports your past
When you rank candidates against each other, the yardstick is implicit, and it tends to become whoever already succeeded here. Same schools, same employers, same resume shape. That feels like a signal, but a lot of it is just similarity to the last hire, which is how a team accidentally makes ten copies of itself and calls it a high bar.
Reading against the role flips the reference point. The question is no longer who looks most like us. It is whether this specific application shows evidence of the specific things the role requires. A candidate who has never worked at a company you recognize can fully evidence every must-have. A candidate with a famous logo can miss most of them. Read against the role, both reads come out honest.
What reading against the role means in practice
- Define the role's must-haves before you read anyone, so the standard is fixed and the same for everyone.
- For each application, check each requirement against the evidence, and mark it evidenced, partial, or unclear.
- Tie every judgment to a verbatim line, so absence of a famous name is never quietly treated as a negative.
That last point matters most. Fairness is not a feeling, it is a discipline: the same rubric, applied the same way, with the evidence shown. When the read is structured like this, two similar candidates get similar summaries whether they applied first or last, and a reviewer can check any conclusion in one click.
Judge the work the application shows, not how much the person reminds you of the last person you hired.Daniel Okafor
Where a human comes back in
Reading against the role gets you a fair, comparable set of first reads. It does not get you a decision, and it should not. Slate surfaces the evidence, the fit, and the gaps to verify, then stops. A person weighs it, asks the follow up questions, and makes the call. The machine is good at reading two hundred applications the same way. The judgment is still yours.