For public sector human resources
No screening, no rating, no ranking, no eligibility or preference determination, and no assessment of any applicant. Those are governed by merit system principles and by veterans’ preference law, they belong to your HR specialists, and an automated hand on any of them is unlawful before it is unwise. What is movable is the waiting between steps, which is where the months actually go.
The elapsed time between deciding to fill a position and somebody starting is measured in months, and almost none of it is anybody deciding anything. It is a position description being classified, a job analysis being agreed, an announcement being drafted and reviewed, a certificate being issued, interviews being scheduled across calendars, a selection being documented, a tentative offer, a background investigation, a suitability determination, and a start date that has to clear a pay period.
Each step has an owner and each owner is reasonable. The delay is between the steps, in the handoffs, and no single person can see the whole clock — the hiring manager sees silence, HR sees a queue, security sees an investigation calendar, and everyone is waiting for somebody who is waiting for somebody.
Candidates leave during that time. The strongest applicants are the ones with alternatives, so the process selects against exactly the people it was trying to attract, and the vacancy is re-announced with the same position description that made it hard to fill the first time.
The hiring manager experiences it as a black box. They asked for a person, they were asked for a position description, and then nothing happened for eleven weeks — during which they made a workload decision based on a start date that was never real.
And the record has to be defensible. A hiring action is reviewable, appealable and occasionally litigated, so every step generates documentation that somebody assembles afterwards under time pressure and files in a place the next reviewer will have to be told about.
No participant in a hiring action can see total elapsed time decomposed by step, so every attempt to speed it up targets whichever step the loudest person is waiting on, which is rarely the one consuming the time.
That is why hiring reform inside an agency so often produces a faster step and an unchanged total. Without the decomposition, the effort lands where attention is, not where the delay is.
The first act is therefore to measure the whole clock, and this requires no authority over anything. For a sample of completed actions: elapsed days per step, wait between steps, rework loops, and the point at which candidates withdrew. Most HR offices can report time-to-hire and almost none can report where it went.
What can then be operated is the logistics band. Confirming a document arrived and is the right one. Scheduling across a panel’s real availability rather than a proposed slot. Reminding a party before a handoff ages rather than after. Keeping the hiring manager informed so they stop having to ask. Assembling the documentation for the file as the action proceeds rather than at the end.
What never moves: rating, ranking, screening, minimum qualification determination, veterans’ preference application, suitability determination, and selection. Those are governed by law and by merit system principles, and they belong to HR specialists and selecting officials. Nothing here reads a résumé to form a view of a person.
Total elapsed time, decomposed by step — measured by days per step and days waiting between steps, against your own completed actions as a baseline.
Wait between handoffs — measured by elapsed days from one owner completing to the next beginning, which is usually the largest component and is rarely measured.
Candidate withdrawals during the process — measured by count of withdrawals and the step at which each occurred.
Interview scheduling rounds per action — measured by count of proposed-and-rejected scheduling attempts before a panel converges.
Hiring manager enquiries asking only for status — measured by count of contacts whose entire content is a status question.
Documentation assembled at the end under pressure — measured by share of file items present at each step versus assembled after selection.
Vacancies re-announced after failing to fill — measured by count of re-announcements, and whether the position description changed between them.
any rating, ranking, screening, minimum qualification determination, veterans’ preference application, suitability determination, or selection. Nothing here reads an application to form a view about a person, and nothing sorts or prioritises candidates by any characteristic or prediction. Those are governed by merit system principles and by law, and they belong to your HR specialists and selecting officials.
It reads the systems your hiring already runs in — the applicant system, the classification library, the panel calendars, the onboarding checklist — through documented interfaces, and writes status and documentation back where your file of record lives.
It does not touch the assessment path. Applications, rating criteria, ratings and certificates are read only to know that a step happened and when, never to form a view about a candidate. If an interface would expose assessment content that is not needed to time a step, that access is not requested.
Announcement and position description assembly draws from your own approved library and template language. Nothing writes a new qualification requirement, because a qualification requirement is a legal statement about who may be considered.
Accessibility is a delivery requirement for every applicant-facing surface, and where an interface does not exist, the limitation is reported rather than covered by driving a screen.
Applicant information belongs to people who applied for a public job and did not choose a vendor. It stays inside your tenancy, on your retention schedule, exportable by you, and it is not used to train anything serving another organisation. Applicant data in particular is never training material, at any tier, under any agreement.
A hiring action is appealable and sometimes litigated, so the record has to be defensible. Every action carries what was done, under which written grant, by which actor, at what time, against which action, with what result — produced as the action proceeds rather than assembled after a selection.
The exclusion from assessment is not a configuration. There is no tier at which this rates, ranks, screens or prioritises a candidate, and that is stated plainly so an HR director does not have to establish it by questioning.
On assurance: an independent SOC 2 Type II attestation is in progress and no report exists yet. Where a jurisdiction carries specific applicant-data obligations, terms are agreed in writing during scoping.
HR policy’s question is whether anything touches the merit path. Nothing does, and the boundary is enumerated rather than described — the administrative band is a written list and everything else routes to a specialist by default.
Counsel and equal employment opportunity should review the measurement definitions, not only the boundary. Anything that changes how quickly different actions move is a question worth testing, and the measurements are defined so your own analysts can test it rather than accept an assurance.
Where a bargaining unit is involved, the arrangement is a change in working conditions for the people whose handoffs are being timed, and it should go through the process that governs that rather than around it. Measuring somebody’s step without telling them is a bad way to begin.
And accessibility is a delivery requirement for applicant-facing surfaces, documented rather than asserted.
A sample of completed hiring actions from the last year, measured retrospectively for elapsed time per step and wait between steps, with no access to assessment content and no authority over anything live.
Retrospective is deliberate: no live candidate is affected, no assessment content is needed, and the finding is checkable against what everybody already remembers. The output is a decomposition almost no HR office has, and it usually contradicts the received view about which step is slow.
Stopping there is a legitimate outcome and often the right one. If the decomposition shows the delay sits in a step outside your control — a background investigation calendar, for instance — then no arrangement inside HR will move it, and knowing that is worth more than a project.
If it continues, the first grant is one logistics act on one series of actions, agreed with any affected bargaining unit before it begins.
Those cases share a mechanism: a system was permitted to screen, rate, rank or prioritise candidates, and the resulting disparity was invisible until it was litigated. Nothing here does any of those, at any tier, and it does not read an application to form a view about a person at all. If the scope of a proposal ever includes assessing candidates, that is the wrong proposal — and the reason is legal before it is technical.
Very possibly true, and the retrospective measurement is how you find out for certain rather than by consensus. If the decomposition shows investigation calendar time dominates, the honest conclusion is that hiring speed is not addressable inside your office and you should stop — which is a finding worth having, because it redirects effort that would otherwise be spent optimising steps that are not the constraint. If the decomposition shows something else dominates, that is worth knowing too.
It would be, if it were introduced as a measurement of individuals, and it should not be. What is timed is the handoff between owners, not a person’s throughput, and where a bargaining unit is involved the change goes through the process that governs working conditions rather than around it. If your specialists cannot see the decomposition themselves, it has been built as surveillance rather than as a tool, and that is a reason to refuse it.
Often both, and the measurement separates them: withdrawals are recorded with the step at which they occurred, so a pattern clustered at a long wait reads differently from one spread evenly. If the data shows withdrawals are not concentrated at any delay, then speed is not your constraint and the honest answer is that this page is not about your problem.