Skip to main content

Documentation Index

Fetch the complete documentation index at: https://nova.dweet.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

Criteria tell Nova what evidence matters for a role. Good criteria make scores useful because they turn a hiring team’s priorities into clear, resume-verifiable screening requirements. Nova generates a first draft using these best practices, so you do not need to start from a blank page or become a criteria expert. Your main job is to add the hiring context Nova cannot infer from the job description, then review whether the draft matches how your team wants to screen.
Criteria are configured per job. You need at least one criterion before Nova can score candidates.

What Criteria Should Do

Good criteria are:
  • Specific to the role: They reflect the actual job, not generic traits.
  • Verifiable: Nova can check them from a resume or application answer.
  • Flexible where appropriate: Equivalent tools, domains, or backgrounds can count when the role allows them.
  • Focused: A typical role works best with 5-8 high-signal criteria.
  • Separated by importance: Must have, Preferred, and Nice to have mean different things.
Criteria should not be scoring instructions, interview questions, or a copy of the full job description.

How Setup Works

From the Jobs page, click Add Criteria or Edit Criteria.
1

Add role context

Include intake notes, hiring manager priorities, non-negotiables, flexibility, deal-breakers, target backgrounds, or application questions that matter. Nova uses this context to draft criteria for you.
2

Confirm priorities

Before the first draft, Nova asks setup questions to confirm priorities, flexibility, deal-breakers, and trade-offs that would change the criteria. It skips questions already answered clearly by the job description, intake notes, or existing application questions.
3

Review the criteria

Check the generated criteria, importance levels, and wording. Most roles only need light edits: tighten anything too broad, soften anything too strict, and remove duplicated or hard-to-verify criteria.
4

Test on sample candidates

Run sample scoring before launch when possible. Look for candidates you expected to review who score too low, candidates with limited relevant evidence who score too high, or repeated confusion around one criterion.

Importance Levels

LevelUse it forHow it affects scoring
Must haveRequirements the role cannot work without.A clear contradiction is a serious concern. Missing evidence creates important uncertainty.
PreferredStrong differentiators that improve confidence.Helps separate strong candidates from acceptable ones.
Nice to haveBonus signals.Adds context, but should not drive the score by itself.
Avoid marking every requirement as Must have. That makes Nova treat minor gaps like major concerns and can hide candidates who are strong overall.

Writing Useful Criteria

Write criteria around evidence Nova can actually see.
WeakBetter
Strong communicatorHas led customer-facing discovery, demos, or stakeholder presentations
Startup mindsetExperience working in early-stage or high-ownership product environments
Good culture fitEvidence of cross-functional collaboration in product, engineering, or customer teams
Knows PythonBackend development experience with Python or a similar production language
Senior enoughHas owned technical design or delivery for complex projects
Use exact tools only when they are truly required. If equivalent experience is acceptable, name the capability instead. If recency matters, say so explicitly. A criterion about a candidate’s current or most recent role is stricter than a criterion that can be satisfied by older experience.

Requirements That Need Application Answers

Some requirements are important but hard to verify from a resume. Examples:
  • Work authorization.
  • Required licenses or certifications.
  • Location or onsite availability.
  • Travel requirements.
  • Willingness to work specific shifts.
  • Specific required experience that is not usually visible on a resume.
For those, add an application question in your ATS. Once the answer is available, Nova can use it as evidence. Do not turn salary expectations, notice period, start date, motivation, or “why us” answers into scoring criteria.
If a requirement cannot be verified from a resume or application answer, it usually should not be a scoring criterion. Keep it for interview evaluation instead.

Deal-Breakers And Exclusions

Use normal Must have criteria for direct eligibility requirements, such as Active RN registration or Right to work in the UK. Use exclusion wording only for objective, verifiable boundaries that are genuinely required for the role. Prefer neutral wording, such as:
  • Exclude candidates whose experience is only academic research, with no production engineering work.
  • Exclude candidates who do not hold the required active license.
  • Exclude candidates who cannot meet a stated location, shift, or work-authorization requirement.
When the evidence clearly verifies an exclusion, Nova treats it as a strong screening rule in the score and assessment. Your team should review the evidence before acting on an exclusion. Avoid exclusions based on inferred motivation, personality, or intent. Nova should not guess that from a resume.

Testing On Sample Candidates

Before starting scoring, test the criteria on a small set of real candidates when possible. Look for:
  • Candidates you expected to review scoring below 7.
  • Candidates with limited relevant evidence scoring above 8.
  • One criterion driving too many concerns.
  • Repeated Partial results caused by unclear wording.
  • Must have criteria that should be Preferred.
When sample scoring shows a pattern, Nova can suggest criteria edits. Review the proposed changes, approve the ones you want, then test again. Sample scoring does not affect your live pipeline.

Editing Criteria

All edits save automatically:
  • Edit text: Click a criterion and update the wording.
  • Change importance: Use the importance dropdown.
  • Delete: Remove criteria that are duplicated, vague, or no longer relevant.
  • Add new: Add a missing requirement.
  • Add from library: Reuse company-approved criteria across jobs.
  • Refine with Nova: Select criteria, describe the change you want, then approve or reject the proposed edits.

Criteria Library

Save reusable criteria to the company library when the same requirement applies across roles. Good library candidates:
  • Work authorization requirements.
  • Location or hybrid work requirements.
  • Role-family requirements that repeat across jobs.
  • Company-specific eligibility rules.
Avoid saving highly role-specific preferences that may be wrong for the next job. Library criteria are reusable starting points. Before adding one to a job, check that the requirement is grounded in that job’s description, intake context, or application questions. For work authorization, travel, onsite availability, licenses, or shift requirements, make sure the job has a matching application question when the resume will not verify it.

Reprocessing After Changes

Criteria changes apply to future scores. Existing scores are not automatically recalculated. If reprocessing is enabled for your workspace, a Nova admin can use Reprocess Applications from the job actions menu when you want previous candidates scored against the updated criteria.
Reprocessing replaces existing Nova scores and assessments for that job.

Best Practices

  • Keep the final list to 5-8 high-signal criteria for most jobs.
  • Use Must have sparingly.
  • Prefer capabilities over brand names or vendor names unless the exact one is required.
  • Split compound requirements when one part could pass and another could fail.
  • Remove duplicated criteria so the same gap is not counted twice.
  • Avoid school prestige, employer prestige, geography, or background proxies unless they are clearly tied to the role. See Bias Testing Methodology for how Nova checks scoring patterns.
  • Test on sample candidates before launch for high-volume or high-stakes roles.
5-8 works well for most jobs. Sparse roles may need fewer. More criteria only help when each one adds a distinct screening signal.
Specific enough to verify, broad enough to allow equivalent evidence. “Backend development with Python, Go, or similar production languages” is usually better than “5 years of Python at a Series B company.”
There is no separate numeric weighting setup. Use Must have, Preferred, and Nice to have to show importance.
No. If Nova cannot verify the trait from a resume or application answer, keep it for interviews instead of scoring.

Understanding Scores

Read and interpret Nova assessments

Enabling Nova

Start scoring on your job postings