# The Hiring-Manager Submittal Standard: Present, Rework, or Hold

*Grade any candidate packet on ten binary criteria and reach the same Present, Rework, or Hold verdict a second grader would, before the hiring manager sees it.*

- Canonical URL: https://www.refolk.ai/guides/hiring-manager-submittal-standard
- Pillar: Recruiting and sourcing
- Format: Standard
- Published: 2026-10-23
- Last reviewed: 2026-10-23
- Reading time: 21 min
- Keywords: what to include in a candidate submittal, how to present candidates to a hiring manager, improve submittal to interview ratio, candidate slate checklist, sourcer to recruiter handoff template, verify candidate employment before submitting

## Key takeaways

- A candidate packet is ready to present only when all ten binary criteria are met, and any single Hold trigger, such as an off-limits employer or a closed requisition, overrides every other strength.
- In Refolk's index, 8% of a 25-profile sample of US sourcer-titled profiles carried unrelated titles, so a headline keyword never counts as proof of a candidate's current role.
- Closed or archived roles account for 21.3% to 27.7% of rejections in two independent datasets, so exclude role-status rejections before you compute a submit-to-interview rate.
- A packet field labelled current salary is itself a pay-history violation in jurisdictions with bans that reach recruiters, so the only lawful comp field is expected pay, with source and date.
- Refolk's index shows 11.8 US Technical Recruiters per titled sourcer, so most handoffs are really self-checks, and that makes blind second grading of a sample necessary.
- If two graders' Cohen's kappa falls below 0.4, the rubric is ambiguous and needs rewriting, and percent agreement plus Gwet's AC1 should be reported alongside it when most packets are Present.

This is the definition of done for a candidate submittal, meaning the point where a sourced, interested candidate's packet goes to the hiring manager. It is written for in-house recruiters, sourcers, and founders who build their own slates. It gives you ten binary criteria, three verdicts with tie-break rules, a handoff template, and a way to prove that two graders on your team reach the same verdict on the same packet.

## What Present, Rework, and Hold mean

Every packet gets exactly one of three verdicts. Present means all ten criteria are met and the packet can go today. Rework means something the recruiter or sourcer can fix is missing. Hold means a fact outside the packet blocks submission, however strong the candidate is.

A **submittal** is the packet a recruiter sends a hiring manager to propose one candidate for interview. A **slate** is the set of submittals for one requisition. The agency Cadre describes each submission as a claim that gets graded within days. This standard has you grade that claim yourself before anyone else does.

### The three verdicts

- **Present.** All ten criteria are met, each with a source and a date. The packet goes to the hiring manager.
- **Rework.** At least one criterion is not met, and the person who built the packet can fix it with more checking or one more conversation with the candidate. The packet goes back to its builder.
- **Hold.** A blocking fact exists that the builder cannot fix: an off-limits employer, a closed or paused requisition, no consent for this role, an identity conflict, or pay, location, or authorization outside what the brief allows. The packet goes to the hiring manager or client as a question, not as a submittal.

### Tie-break rules

These rules let two graders reach the same verdict on the same packet.

1. **Hold beats Rework, and Rework beats Present.** One Hold trigger makes the packet Hold, whatever else it contains.
2. **Undated means not met.** A check with no date has not been done, for the purposes of this standard.
3. **Unverifiable means not met.** The exception is a brief that explicitly accepts the claim as unverifiable.
4. **"None found" counts as met only when it names the source searched and the date.** A bare "none found" is Rework.
5. **A marked gap on a must-have means not met for Present.** The exception is a brief where the hiring manager agreed in advance to see gap candidates. Then the gap has to appear in the packet's summary line.
6. **Conflicting employer or title is Rework when one call to the candidate can settle it.** If the conflict is about identity itself, such as a name, history, or profile that does not hold together, it is Hold.

#### Where a packet lands

Horizontal axis runs from Evidence has gaps to Every must-have cited. Vertical axis runs from No blocking fact to Blocking fact present.

| Quadrant | What it means |
| --- | --- |
| Hold | Resolve the blocker with the hiring manager before spending any more effort on evidence |
| Hold | Strong candidate with a hard stop - escalate the blocker and do not submit |
| Rework | Return to the builder with the specific rows that failed |
| Present | Submit today with the evidence map attached |

*Blocking risk overrides evidence quality, so a fully evidenced packet with a hard stop is still a Hold.*

## The ten criteria a packet is graded on

A packet is graded on six person-level criteria, three employer-level criteria, and one role-level criterion. Each one is scored met or not met, with no partial credit. Every criterion states what it proves and what it looks like when it lies, because a check that cannot fail is not a check.

#### Layers of a submittal packet

1. **Consent and role status** - The candidate agreed to this role, and the requisition is still open
2. **Employer context** - Off-limits status, WARN and acquisition events, pay-calendar timing
3. **Logistics** - Expected pay, notice period, location, work authorization
4. **Evidence map** - One cited proof per numbered must-have, and experience inside the range
5. **Identity and current employment** - Employer and title confirmed on two surfaces and by the candidate

*Each layer rests on the one beneath it, so a packet with an unverified employer cannot be rescued by strong logistics.*

### Person-level criteria

| # | Criterion | Proves | Lies when |
| --- | --- | --- | --- |
| C1 | Current employer and title verified | The candidate is who and where they say | Headline keyword or stale employer field |
| C2 | Evidence map complete | Each must-have has a citable proof | Evidence is generic, not tied to the row |
| C3 | Experience inside brief's range | Neither under- nor over-levelled | Range never written into the brief |
| C4 | Expected pay confirmed | Offer is feasible within band | Field actually records pay history |
| C5 | Notice, location, authorization | Start is feasible | Recruiter inferred it from the profile |
| C6 | Contact and consent recorded | Candidate agreed to this submission | Consent was for a different role |

### Employer- and role-level criteria

| # | Criterion | Proves | Lies when |
| --- | --- | --- | --- |
| C7 | Off-limits clear | Submission breaches no client contract | List built from competitor deals |
| C8 | Employer events dated | Layoff or acquisition context known | "None found" below WARN thresholds |
| C9 | Pay-calendar timing recorded | Vesting or bonus timing won't derail an offer | Assumed rather than asked |
| C10 | Requisition confirmed open | A rejection here would be about the person | Status checked weeks before submit |

### Why C1 needs two surfaces and the candidate

C1 is met only when the current employer and title appear on two independent surfaces and the candidate has confirmed them directly, with the date recorded. Independent surfaces include the professional profile, the company's own site, the candidate's public GitHub activity, and conference or publication bylines. Two copies of the same profile do not count.

**8% - Title noise in a sample of US sourcer-titled profiles**

In Refolk's index, 2 of 25 profiles returned for sourcer titles held unrelated roles, including a company president.

Title noise is exactly what C1 is built to catch. A search for a title brings back people whose headline contains the word, and that is not proof they hold the role now. [Refolk](/) returns matches across LinkedIn, GitHub, and the open web together, so the second surface is usually already in front of you, not a separate search.

Fraud raises the stakes. A background-screening provider's survey of 3,000 US managers found that 60% had uncovered a candidate who misrepresented their experience or qualifications, and another 13% suspected it but could not prove it.

| Measure (US managers) | Earlier edition | Later edition |
| --- | --- | --- |
| Suspected AI misrepresentation | 59% | 67% |
| Interviewed a fake identity | 31% | 40% |
| Extremely confident process catches fraud | 19% | 23% |

The same survey series also reports HR fraud training rates, but its two editions give conflicting figures for the earlier year, so I don't rely on that row. The point that survives is that fewer than a quarter of managers are extremely confident their process would catch a fraudulent candidate. Reviewer judgment has not earned trust here. A recorded verification state (verified, conflicting, or unverifiable) has. I found no published method for checking candidates against public records before submission, so treat C1 as your team's own control and audit it.

### Why C3 is a separate criterion from C2

C3 exists because experience calibration is the biggest merit-based reason for rejection, and it fails in both directions. In a recruiting-platform study of more than 500,000 merit-based decisions coded with a 25-value taxonomy, lack of experience led at 16.6% and too much experience ranked fourth at 12.2%. Both can be checked before submission, but only if the brief states a tenure and scope range. If it doesn't, C3 cannot be graded, and the fix belongs in the brief, not in the packet. Treat these as vendor figures.

## Which fields a submittal must never contain

A submittal must never contain the candidate's current or past pay. The only lawful comp field is expected pay, recorded with its source and date. In jurisdictions with pay-history bans that reach recruiters, a "current salary" field in the template is a violation by itself, whoever fills it in.

- **United States.** A growing number of states, 18 plus Washington, D.C. at the statute review I relied on, restrict employers from asking applicants about pay history. California's ban covers asking "orally or in writing, personally or through an agent", so it reaches recruiters. Virginia has extended its ban to private employers. The list keeps changing, so re-check it against current statute before you adopt this policy.
- **European Union.** Article 5(2) of the EU Pay Transparency Directive bans employers from asking candidates about current or previous salary, whether directly, through application forms, or through third parties such as recruiters. National transposition has been uneven, and only four of the 27 member states met the deadline. Check the local implementing law for each country you hire in.
- **Work authorization and protected characteristics.** I have not verified what you may lawfully ask or record about these in each jurisdiction. Before C5 goes live, have counsel confirm the exact wording you may use for authorization, and keep protected characteristics out of the packet entirely.

> **Rule:** Expected pay, never pay history
>
> The packet's comp field reads "expected pay", with the candidate as source and a date. Delete any current-salary or salary-history field from the template itself, not just from individual packets.

## Who owns each criterion in the sourcer-to-recruiter handoff

The sourcer owns identity and evidence (C1 to C3). The recruiter owns employer context, logistics, consent, and role status (C4 to C10). When one person does both jobs, which is the common case, the handoff template becomes a self-check, and that is the reason a second grader matters.

| Segment | Profiles | Top employers in 25-profile sample |
| --- | --- | --- |
| Technical Recruiter, US | 21,396 | K2 Partnering Solutions (3), Experis (2) |
| Technical Recruiter, UK | 716 | Monzo (3), Sharp Brains (3), Meta (2), Capgemini (2) |
| Sourcer titles, US | 1,819 | Rippling, MongoDB, American Express, Verkada (1 each) |

Source: Refolk's index. The employer counts come from a 25-profile sample only, so they show texture, not market share.

**11.8 - US Technical Recruiters per titled sourcer**

Derived from Refolk's index of professional profiles - most recruiters have no dedicated sourcer to hand off to.

The same index gives 29.9 US Technical Recruiters for every UK one. That ratio says nothing about quality. What it does mean is that a standard written for a large US team has to work just as well for a recruiter in a small UK market grading their own packets alone.

**Sourcer-to-recruiter handoff packet**

```
REQ: [requisition ID] | Brief version: [v#, date locked]
CANDIDATE: [full name] | Primary profile: [link] | Second surface: [link]

C1 CURRENT EMPLOYER AND TITLE
Employer: [name] | Title: [title] | Since: [month/year]
Surface 1: [source, date checked] | Surface 2: [source, date checked]
Candidate confirmed: [yes/no, channel, date]
State: verified / conflicting / unverifiable

C2 EVIDENCE MAP (one row per numbered must-have)
MH1 [must-have text]: [specific evidence + link] OR GAP
MH2 [must-have text]: [specific evidence + link] OR GAP
MH3 [must-have text]: [specific evidence + link] OR GAP

C3 EXPERIENCE VS BRIEF RANGE
Brief range: [tenure / scope] | Candidate: [tenure / scope] | Inside / under / over

C4 EXPECTED PAY (never pay history): [amount or range, candidate-stated, date]
C5 NOTICE: [period, date] | LOCATION: [city, on-site/hybrid/remote fit, date] | AUTHORIZATION: [counsel-approved wording, date]
C6 CONSENT: [channel] | [date] | Agreed to be submitted for: [role, company]
C7 OFF-LIMITS: [clear / hit] | Contract cited: [client agreement reference] | [date checked]
C8 EMPLOYER EVENTS: WARN [result + source searched + date] | Acquisition [result + source searched + date]
C9 PAY-CALENDAR TIMING: [candidate-stated vesting or bonus timing, or "not disclosed", date]
C10 REQ STATUS: Open, confirmed with [hiring manager] on [date]

VERDICT: Present / Rework / Hold | Failed criteria: [list] | Grader: [name] | Date: [date]
SUMMARY LINE FOR HIRING MANAGER: [one sentence, naming any pre-agreed gap]
```

*One packet per candidate per requisition. Any line without a source and a date grades as not met.*

## How to grade a packet, step by step

Grade every packet in the same seven steps, in the same order, so the verdict comes from the record and not from the grader's impression. No public timing data exists for any of these steps. Measure your own durations before you set service levels.

#### Submittal grading procedure

1. **Lock the brief** - Recruiter and hiring manager write down the numbered must-haves, the tenure and scope range, the pay band, and the off-limits companies drawn from client contracts. Done means a dated brief where every must-have has a number.
2. **Verify identity and current employment** - The sourcer cross-checks title, employer, and tenure across two independent surfaces and the candidate's own dated statement. Done means each claim is marked verified, conflicting, or unverifiable.
3. **Map evidence to each must-have** - The sourcer cites one specific piece of evidence per numbered must-have, or marks the row as a gap. Done means no blank rows and an explicit inside, under, or over call on experience.
4. **Check employer context** - The recruiter checks the off-limits list, public WARN filings, and acquisition news, and asks the candidate about pay-calendar timing. Done means every item has a dated source or a dated none-found naming what was searched.
5. **Confirm logistics** - The recruiter confirms expected pay (never pay history), notice period, location, and authorization directly with the candidate. Done means each item came from the candidate and carries a date.
6. **Record contact and consent state** - The recruiter records the channel, the date, and the exact role and company the candidate agreed to be submitted for, and reconfirms the requisition is open. Done means consent names this requisition.
7. **Give the verdict and second-grade a sample** - Apply the tie-break rules to reach Present, Rework, or Hold, then have a second grader score a blind sample. Done means the verdict is logged and agreement is computed for the cycle.

### Notes on the order

- **No source fixes whether employer checks come before or after evidence mapping.** I map evidence first because a candidate who fails the must-haves doesn't need employer checks. The off-limits lookup is cheap, though, and it is a hard stop. If your list lives in a spreadsheet you can search in seconds, run it straight after step 2.
- **Screening is not submitting.** Recruiterflow treats screen-to-submission as a separate, earlier step. This standard begins once a candidate has been screened and is interested.
- **WARN filings date the event.** The US WARN Act requires at least 60 calendar days' advance written notice of a plant closing or a mass layoff affecting 50 or more employees at a single site. Notices are public, state by state, and some states publish them as open datasets. When you find a filing, record the filing date and the site, not just "layoffs".

When C8 turns up a layoff, the candidate's packet usually gets stronger, because the recruiter can explain why someone would move now. The same check works in reverse for sourcing, when you want people at employers that have just filed. Refolk takes that as a plain-English request:

Ask me this: `Senior backend engineers in Texas whose current employer filed a WARN notice in the last 90 days` - [run the search](https://www.refolk.ai/start?q=Senior%20backend%20engineers%20in%20Texas%20whose%20current%20employer%20filed%20a%20WARN%20notice%20in%20the%20last%2090%20days).

*Returns named engineers with current employer and title, which you then verify on a second surface and date under C1 and C8.*

## How submittal grading goes wrong

Submittal standards fail in predictable ways. Metric definitions drift, title and employer fields lie, role closures get blamed on candidates, comp fields leak pay history, off-limits lists get built illegally, and "none found" gets read as "nothing happened". Each failure below has a check you can write into policy.

### 1. Ratio definition drift

The same team can report 4:1 or 25% for the same quarter depending on which formula it uses. Counting a phone screen as an "interview" inflates the rate further. One glossary defines the ratio as submittals divided by interviews scheduled. Recruiterflow uses the share of submitted candidates the client agrees to interview. **Check:** write the name of the interview stage, as it appears in your ATS, into the policy, and report one formula only.

### 2. Title match noise

In Refolk's index, 8% of a sample of US sourcer-titled profiles held unrelated roles. A headline keyword is evidence of a headline, nothing more. **Check:** C1 requires two independent surfaces plus the candidate's dated confirmation.

### 3. Stale or odd employer field

In a sample of UK Technical Recruiter profiles, a chemicals company appeared as a top employer. That kind of field looks like a verified current employer and isn't one. **Check:** match the employer against the company's own domain, and get the candidate to confirm it with a date.

### 4. Role closed, counted as candidate fail

Closed or archived jobs made up 21.3% of rejections in the recruiting-platform study, and the staffing firm VA Masters found 27.7% of rejections happened because the role closed or was put on hold. If you leave those in, your submit-to-interview audit punishes recruiters for requisition churn. **Check:** C10 at submission, and exclude role-status rejections before you compute the rate.

### 5. Comp field leaks pay history

A template with a "current salary" field breaks the EU Article 5(2) ban and US state bans, including when an agent collects the figure. **Check:** the field reads "expected pay", with source and date, and the template has no other comp field.

### 6. Kappa paradox

When most packets are graded Present, Cohen's kappa can stay low even with 95% raw agreement. That makes a working rubric look broken. **Check:** report percent agreement and Gwet's AC1 alongside kappa.

### 7. Off-limits list built from competitor agreements

Joint DOJ/FTC antitrust guidelines on worker issues state that agreements not to hire or solicit from other companies are per se illegal. A legitimate off-limits restriction comes from your own client or service agreement, never from a deal between competing employers. **Check:** every entry on the off-limits list cites the client contract it comes from.

> **Watch out:** "None found" on WARN is not "no layoff"
>
> Employers below the WARN size and headcount thresholds file no notice, so an empty search proves only that nothing was filed. Record C8 as "no filing found in [source], [date]" and still ask the candidate about recent team changes.

### 8. WARN false negative

This is the warning above as a rule. An empty WARN search is a true negative only for employers the Act covers. For small startups, the candidate's own account, with its date, is the main evidence. Grade C8 as met on the documented search, not on the absence of news.

> A packet that cannot fail any of its checks is a formatting exercise, not a submittal.

## How to audit the standard against your submit-to-interview rate

The standard is working when your submit-to-interview rate rises against your own baseline and two graders reach the same verdict on the same packets. Published benchmarks are weak context, because none defines "interview" and none splits in-house teams from agencies.

| Source | Figure | Population | Definition |
| --- | --- | --- | --- |
| Cadre, citing unnamed "HR data" | 46% average | Unspecified | Client wants to interview |
| Cadre | 3:1 to 5:1 "healthy" | Firms | Submittals per interview |
| Cadre, own rate | 84% | One agency, software engineers | Client wants to interview |
| eTeki | 3:1 or better; 6:1 to 8:1 in some industries | Staffing | Submitted per interview given |

Read the table with care. The 46% figure has no named origin. The 84% is one agency's own published rate on its own engagements. The eTeki figures cover staffing only and are several years old. Nobody has publicly established a split between in-house and agency rates. Use your pre-standard quarter as the baseline.

### Running the audit

1. **Fix the definition.** Write down the formula and the interview stage name, and don't change either mid-audit.
2. **Code rejection reasons.** Use your ATS's structured rejection reasons. Greenhouse, for example, documents a rejection reasons report. Code role-status rejections (closed, on hold, archived) separately.
3. **Exclude role-status rejections**, then compute the rate.
4. **Separate non-response.** The same recruiting-platform study coded 28.7% of reasoned rejections as the candidate never responding to outreach. That happens before submittal and should never show up in a submittal metric. If it does, C6 is being skipped.
5. **Break failures down by criterion.** For every hiring-manager rejection of a packet graded Present, note which criterion the stated reason maps to. A cluster on C3 means the brief's range is wrong. A cluster on C1 means verification is too thin.

### Measuring grader agreement

**Cohen's kappa** measures agreement between two graders, corrected for the agreement you would expect by chance. Published bands treat 0.41 to 0.60 as moderate and 0.61 to 0.80 as substantial. Below 0.4, the rubric is ambiguous and needs revising. **Gwet's AC1** is an alternative coefficient that doesn't collapse when one verdict dominates, and it is the answer to the kappa paradox above. The sources give no minimum sample size. Start with as many packets as a second grader can score blind each cycle, and grow the sample if the coefficients swing from cycle to cycle. There is a public guide on choosing an inter-rater coefficient if your team wants to go further.

**Second-grader agreement log**

```
Cycle: [dates] | Sample size: [n] | Interview stage counted: [ATS stage name]

Packet ID | Grader A verdict | Grader B verdict | Criteria scored differently | Resolution
[ID] | Present / Rework / Hold | Present / Rework / Hold | [e.g. C3, C8] | [rule clarified or grader error]

Percent agreement: [matching verdicts / n]
Cohen's kappa: [value] | Band: below 0.4 revise / 0.41 to 0.60 moderate / 0.61 to 0.80 substantial
Gwet's AC1: [value]
Rubric change this cycle: [criterion, old wording, new wording] or none
```

*Grader B scores blind from the packet alone. Log every sampled packet, then compute the three summary figures per cycle.*

## Pre-submission checklist

Run this immediately before you send any packet. If any item is unchecked, the packet is Rework or Hold, and the tie-break rules decide which.

#### Before any packet reaches the hiring manager

- [ ] The brief is locked, dated, and every must-have carries a number
- [ ] Current employer and title appear on two independent surfaces and the candidate confirmed both, with a date
- [ ] Every numbered must-have has one cited piece of evidence or is marked as a gap
- [ ] Experience is recorded as inside, under, or over the brief's written range
- [ ] Expected pay came from the candidate and is dated, and no pay-history field exists in the packet
- [ ] Notice period, location, and authorization came from the candidate and are dated
- [ ] Consent names this exact role and company, with channel and date
- [ ] The employer is checked against an off-limits list where every entry cites a client contract
- [ ] WARN and acquisition checks each show a result, the source searched, and the date
- [ ] Pay-calendar timing is recorded as the candidate stated it, or as not disclosed
- [ ] The hiring manager confirmed the requisition is open on the day of submission
- [ ] Any pre-agreed gap is named in the summary line

## Keeping the standard current

A submittal standard goes stale in three places: the law behind the comp field, the public sources behind the employer checks, and the wording graders disagree about. Put each on a recurring review and log the changes.

- **Pay-history law.** The US count of states with bans and the EU's national transpositions both keep moving. On a fixed cadence, re-check each jurisdiction you hire in against current statute, and have counsel sign off on the C5 authorization wording at the same time.
- **WARN sources.** State labor departments publish notices in different formats, and some offer open datasets. Keep a list of which state sources you search, so a "none found" under C8 always names a real source.
- **Off-limits list.** Re-check every entry against its cited client contract when contracts renew. Remove any entry that has no contract behind it.
- **Rubric wording.** When kappa falls below 0.4, or one criterion keeps showing up in the "scored differently" column of the agreement log, rewrite that criterion's met-when wording before the next cycle and log the old and new text.
- **Replacement sourcing.** When Hold verdicts cluster on one off-limits employer or one closed requisition, rebuild the candidate pool from open employers instead of re-grading the same packets. Refolk takes that as a plain request, for example the same role at named companies that are not on the list, and the new packets go through the same ten criteria.

If you run this for a full audit cycle and the submit-to-interview rate doesn't move against your baseline, check the brief before the packets. A brief without a written experience range or numbered must-haves gives C2 and C3 nothing to grade against.

## Frequently asked questions

### What should a candidate submittal include?

A submittal should include verified current employer and title, one cited piece of evidence per numbered must-have in the brief, employer context (off-limits status, WARN and acquisition checks, pay-calendar timing), expected pay, notice, location, authorization, and a record of what the candidate consented to be submitted for. Every item needs a source and a date. Formatting matters far less than whether each claim can be checked by the hiring manager.

### What is a good submittal-to-interview ratio?

Published benchmarks cluster around 3:1 to 5:1 submittals per interview, and one agency cites an unnamed industry average of about 46% of submittals reaching interview. None of these sources defines what counts as an interview or splits in-house from agency teams, so treat them as loose context. Your own baseline, measured with a fixed interview definition and role-status rejections excluded, is the number to beat.

### Can a recruiter ask for a candidate's current salary for the submittal?

In many places, no. A count of US states plus Washington, D.C. restrict employers from asking about pay history, California's ban reaches requests made through an agent, and Article 5(2) of the EU Pay Transparency Directive covers third parties such as recruiters. Ask for expected pay instead, record it with source and date, and keep any current-salary field out of the packet template entirely.

### How do I verify a candidate's current employer before submitting?

Confirm the employer and title on two independent public surfaces, such as the professional profile and the company's own site or the candidate's GitHub activity, then get the candidate to confirm it directly and date that confirmation. If the surfaces disagree, mark the claim conflicting and resolve it before submitting. A headline keyword or one stale employer field does not count as verification.

### What is the difference between Rework and Hold?

Rework means the packet is missing something the recruiter or sourcer can fix, such as a blank evidence row, an undated check, or an unconfirmed notice period. Hold means a fact outside the packet blocks submission: an off-limits employer, a closed requisition, no consent for this role, or an identity conflict. Rework goes back to the person who built the packet. Hold goes to the hiring manager or client for a decision.

### How do I check that two recruiters grade packets the same way?

Have a second grader score a blind sample of packets each cycle and compute percent agreement and Cohen's kappa. Kappa of 0.41 to 0.60 is moderate and 0.61 to 0.80 substantial, while below 0.4 means the rubric is ambiguous. Because most packets often land on Present, which depresses kappa, also report Gwet's AC1.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/hiring-manager-submittal-standard*