← All insights

Hiring ops

Résumé parsing is a solved problem. Ranking isn't.

Extracting skills from a PDF is table stakes. Weighing them against a real brief is where an ATS earns its keep.

4 min read

Every applicant tracking system can pull a list of skills out of a PDF. It has been able to for years. If a vendor leads with parsing accuracy, they are describing the part of the problem that stopped being hard a long time ago.

The hard part starts after parsing. You have three hundred structured profiles. Which ten should a human read first?

Why keyword matching fails

A profile that mentions distributed systems once, in a bullet about a course, matches the same keyword as one from someone who has spent eight years designing them. A synonym the posting did not use scores zero. A capability that was central to the candidate's last role but phrased differently is invisible.

Keyword matching measures overlap between two documents. It does not measure whether the person can do the job.

What ranking should actually weigh

Depth: not whether a skill appears, but how much evidence supports it. Relevance: how closely that evidence maps to this brief, not to the field in general. Recency: a capability exercised last year counts for more than one from a decade ago. Trajectory: whether the candidate's responsibility has been growing.

Each of those needs a weight, and the weights come from the brief. A role that needs a safe pair of hands and a role that needs a pioneer should rank the same pool differently.

Ranking against the brief, not the industry

The brief is the anchor. Before screening starts, it is translated into weighted criteria: what is essential, what is valuable, what is nice to have. Every application is scored against those criteria, with the evidence recorded, and the ranking follows from the scores.

That means the same candidate can rank first for one role and outside the top fifty for a similar-sounding one. That is not inconsistency. That is the system doing its job.

Why a person still reads the top of the list

Because ranking narrows; it does not decide. The purpose of a good ranking is to make sure the ten profiles a human reads are the ten worth reading. The purpose of a bad one is, apparently, to save the human from reading anything, which is how strong candidates disappear.

We rank so that your attention lands in the right place. What you do with it is still up to you.

Hiring for depth?

Submit a brief and receive a ranked shortlist within 48 hours.