The words, and what they actually mean.

An internally approved open role, with an owner and, where your process requires it, a sign-off before it can be published.

A requisition is not a job posting. The requisition is the authorised headcount; the posting is its advertisement. Indian agency recruiters usually call the same thing a requirement or a mandate.

The document describing a role. On its own it is prose, which is why nobody can be scored against it.

Parsed into a weighted set of must-haves it becomes something measurable. A good parse keeps 5+ years React at five years rather than quietly softening it into a preference, and reads Rs. 18,00,000 as ₹18 lakh rather than eighteen thousand.

The weighted, editable list of must-haves, should-haves and nice-to-haves derived from a JD.

Every candidate is scored against this and nothing else, which is why a human approves it before the role goes live. A bad spec poisons every score beneath it.

A requirement treated as a dealbreaker. A candidate who misses one has their score capped, or is knocked out entirely.

Marking a should-have as a must-have is the most common cause of why is everyone scoring so low.

Sourcing is finding people who are not yet in front of you: board search, Boolean, outreach, referrals, headhunting.

Screening is filtering people you already have against the role's criteria. Vetting is verifying claims: skills testing, references, background checks.

Re-scoring your existing database against a new role is screening, not sourcing. Any vendor who calls it sourcing is telling you something about themselves.

A candidate's fit against one role's requirement spec, expressed 0 to 100. In CheckedIn it is arithmetic over seven factors: skills, experience, education, location, salary, notice period and employment type.

It is not an AI opinion. AI parses the JD and the CV; the scoring itself is fixed arithmetic. That is precisely why the same shortlist ranks identically tomorrow, and why re-ranking your whole database after a correction costs nothing.

The line-by-line breakdown behind a match score: each requirement, its weight, how it matched, the evidence in the CV, how recent that evidence is, and what it contributed to the total.

The most useful column is matched via. It is the difference between they listed Postgres and they listed Postgres, the job asked for PostgreSQL, and those are a known alias.

Because the explanation is the same arithmetic as the number, the two cannot disagree.

A result from an assessment the candidate actually sat, as opposed to a claim on their CV.

In CheckedIn it replaces the claim inside the ranking rather than sitting politely beside it. List React, score 0.42 on the React section, and your React match becomes 0.42. You get no credit for having typed the word.

It can never be edited. A re-grade is a new assessment, and both stay visible.

A condition that holds a candidate at a pipeline stage: a minimum measured score, required interview feedback, or an approval.

A gate is only meaningful if it is enforced in the core of the system rather than by hiding a button. A candidate held at a gate is the gate working, not a bug.

An agency sending a named candidate to a client's role.

A submission is not an application. An application is candidate-initiated and inbound. A submission is an agency putting a named person in front of a client, and it is the moment candidate ownership is decided and a fee becomes possible. Different actor, different consequence.

Which agency holds the right to a fee on a candidate. Decided once, at the first submission, and sticky for the agreed window.

The submission time is stamped by the server, not by the submitter, because submission order decides money and submitters have an obvious incentive.

This is not a Right to Represent. An RTR is a candidate-signed US document authorising one agency to present them to one named client. Ownership here is the client's own attribution record.

How long first-submitter ownership lasts. Ninety days is common, but it is whatever your MOU says.

Nothing automatic can reassign ownership inside the window. Any transfer after it expires is an explicit, recorded action rather than a silent one.

Two agencies submitting the same person for the same role.

For the second agency this is a total loss: the sourcing cost was already spent and the fee goes to someone else. It is the single largest avoidable margin leak on an agency desk.

Handled well, the second submitter is told plainly who owns the candidate and until when, which turns an argument three months later into a commercial decision today.

Being formally added to a client's approved list of recruitment partners, with agreed fee terms.

In practice: a connection the client approves, then access granted to specific roles rather than to everything by default.

Empanelment is not onboarding. Onboarding is what happens to a new joiner.

The commercial agreement with one agency: fee percentage, replacement window, payment terms, ownership window.

Fee terms should be versioned, so renegotiating next quarter does not retroactively change what you owe on a placement made under the old ones.

India anchor: 8.33% of annual CTC is one month's salary (one twelfth). Typical volume work runs 8 to 12%, niche and senior 12 to 15%, leadership 15 to 20%, retained or executive search 20 to 33%.

A candidate who joined. Not an offer, and not an acceptance.

Joining is what the fee waits for, and joining is what raises the invoice. Cautious agency owners do not count a placement until the replacement window has closed.

Keep onboarding for the aftermath: documentation, induction, IT provisioning, BGV closure.

A candidate who accepted an offer and then did not join.

This is why offer acceptance rate and offer-to-join rate are two different numbers that must never be quoted as one. In India, ninety-day notice periods and offer shopping make the gap between them the normal case rather than the exception.

Time-to-fill runs from the requisition opening to the role being filled, so it includes approval delay and hiring-manager latency. It is owned by the organisation.

Time-to-hire runs from a candidate entering the process to their offer being accepted. It is a per-candidate measure of process efficiency.

They are not interchangeable, and any number quoted without its endpoints is not a number. The same applies to fill rate without its denominator.

How many submissions it takes to produce one client interview.

This is the number that decides an agency's cost per placement, and the one a client quietly uses to decide whether you stay on the panel.

Worked example: a recruiter costing ₹45,000 a month who makes 30 submissions has a cost per submittal of ₹1,500. At one interview per ten submissions, every interview costs ₹15,000. Move it to one in four and the same interview costs ₹6,000, without hiring anyone.

On contract staffing, markup is (bill rate minus pay rate) divided by the pay rate. Gross margin is the same numerator divided by the bill rate.

A 1.5× markup is a 33.3% gross margin, not 50%. Then employer PF, ESI and gratuity accrual come out of the spread, so the realised number is lower again.

A queue of work the system has routed to a specific person, rather than a notification feed you scan and hope you did not miss anything.

The properties that make it work: one task to one owner, deep-linked to the exact screen with the action loaded, self-closing when the work is done, and raised once rather than every time a background scan runs. A queue that re-raises the same item every fifteen minutes trains everyone to stop reading it.

The system of record for requisitions, candidates, applications and pipeline against currently open roles.

A recruitment CRM is a different thing: the client and prospect relationship system, with business-development pipeline, deal stages and outreach sequences.

CheckedIn is an ATS with an agency-relationship and fee-attribution layer on top. It has no BD pipeline and no outreach sequences, so it is not a CRM.