The good applicant to the wrong job
Inbound CVs get read against the job they were sent for, by the consultant who owns that job. If the fit is wrong, the CV is rejected or parked, and the three other roles it would have suited never hear about it. That is not carelessness. With dozens of roles open across a team, nobody holds every spec in their head, and nobody has the time to check one applicant against all of them.
This play is the inverse of find the candidates already in your ATS. That one takes a single role to many candidates. This one takes a single candidate to every live role, and most of the work is a free read of your own system.
What you get back
A ranked list of your live roles for this candidate: the role, the client, the score, the lines from the CV or profile that support it, and the must-haves that are missing. Roles that were ruled out are listed with the reason, so you can see that they were considered. If the person is already in your database, their history with you sits alongside. Any change to the ATS is a proposal waiting for your approval.
Variations worth knowing
Nothing fits. A strong candidate with no live role is not a dead end. The next step is to spec the candidate out to companies that should want them.
A day's applicants at once. The same scoring runs across a batch of inbound CVs. The work grows with candidates multiplied by roles, so the pilot and the cost check matter more, not less.
A profile instead of a CV. A LinkedIn URL works as the starting point. The dated work history is looked up first, because a current title and employer on their own cannot evidence a years-of-experience requirement.
Where this goes wrong
Thin job records. A job that is a title and two lines cannot be scored against. The play says which roles it could not assess properly instead of guessing at them.
Open jobs that are not really open. A role nobody closed in the ATS gets scored as live. That is a data hygiene problem, and this play keeps surfacing it until it is fixed.
Scanned and designed CVs. A CV with no text layer, such as a scan or a design-tool export, is read as an image instead. It works, but check the parsed profile on those before you trust the scores.
Rounding up. A near miss presented as a fit costs you credibility with the client. The missing must-have stays visible and holds the score down, which is the behaviour you want even when it is inconvenient.
Protected attributes. Age, gender, marital status and similar details are never scored on, whatever a CV or a custom field happens to contain.
