Glossary

Software employers use to manage candidates through hiring pipelines — sourcing, screening, interviews, offers, and onboarding. ZhiYin functions as an ATS for participating employers.

A query language combining keywords with AND, OR, NOT, parentheses, and quotes to narrow candidate searches. Often used in talent sourcing.

The end-to-end quality of a candidate’s journey through a hiring process: clarity, responsiveness, fairness, and respect for their time. High candidate experience improves offer-acceptance rates and employer reputation.

Outreach to passive candidates (not actively applying) who match a role. Effective cold outreach is personalized, concise, and respects opt-out.

Total internal plus external recruiting spend divided by the number of hires in a period. SHRM/ANSI standard formula. US average sits near $4,700 (SHRM 2022 Talent Acquisition Benchmarking Report), substantially higher for executive and engineering roles.

Outreach designed to widen the top of the funnel to underrepresented candidates while keeping selection criteria identical for every applicant. Distinct from quotas; lawful in most jurisdictions only when assessment standards remain neutral.

A structured incentive system where current employees refer candidates from their network in exchange for a bonus when the referral is hired and stays past a vesting threshold. Referral hires typically have higher 1-year retention than other sources.

Accepted offers divided by offers extended over a period. SHRM benchmarks typical acceptance rates around 85–90%; rates below 70% usually flag compensation, candidate-experience, or competing-offer problems worth investigating.

A formal written employment offer covering title, start date, compensation, equity (if any), benefits, location, and contingencies. Binding once countersigned. On ZhiYin, offers can be tracked and compared inside /candidate/offers.

The first 30-90 days after a placement: paperwork, equipment setup, role context, team introductions, and a 30/60/90 plan. Strong onboarding correlates with higher 6-month retention and faster ramp-to-productivity.

A professional who is employed and not actively job-hunting, but open to the right opportunity. Most senior hires come from this pool — reach them through warm intros, recruiter outreach, and reputation, not job-board postings.

The sequence of stages a candidate moves through for a role — typically Applied → Reviewing → Interviewing → Offer → Hired.

A confirmed hire. On ZhiYin, a placement is recorded when an employer reports a hire and the candidate confirms, which triggers the platform fee.

A composite measure of how well new hires perform after joining. Common inputs: ramp time, first-year retention, manager rating at 90/180/365 days, and performance review score. LinkedIn’s 2024 Future of Recruiting report ranks it the #1 metric recruiting leaders track.

A late-stage verification call where the employer speaks with people who have worked with the candidate (manager, peer, direct report) to confirm scope of impact and surface red flags. ZhiYin logs reference-check attempts to the audit trail so candidates can see who was contacted.

A formally approved request to hire for a specific headcount slot. Carries budget, role scope, and approval chain. A job posting is the public-facing artifact of an opened requisition.

The channel that produced a hired candidate — referrals, job boards, sourcing outreach, agencies, careers site. Referrals consistently lead in retention and quality across industry studies (Jobvite, Lever).

A curated set of candidates an employer maintains for current or future roles, segmented by skills, seniority, or interest signals.

Days between a job opening and the accepted offer. Tracked per role and per recruiter; used to forecast hiring capacity.

Days from a candidate’s application to their accepted offer. Different from time-to-fill: it measures pipeline efficiency for a specific candidate, not how long a job sits open.

An interviewer outside the hiring team trained to maintain a consistent hiring bar across the company. Originated at Amazon and widely adopted.

An interview format that asks for concrete past examples ("Tell me about a time…") to predict future behavior. STAR framework (Situation, Task, Action, Result) is the standard answer structure.

A meeting where interviewers align on scoring rubrics and discuss a candidate’s signals together. Reduces individual bias and rating drift.

A post-onsite meeting where every interviewer presents their evidence-based recommendation (hire / no-hire / lean) before voting. Run well, it surfaces blind spots; run badly, it lets the loudest voice anchor the decision.

An early-round conversation (usually 30 minutes) with the role's direct manager. Confirms fit on motivation, scope expectations, and pay range before investing in panel rounds. Failing it is the most common silent rejection — treat it as load-bearing.

The scoring sheet interviewers use after a loop — typically dimensions like technical depth, problem solving, collaboration, ownership, scored on a fixed scale. Reduces halo effect and rater drift versus free-form notes.

A coordinated series of back-to-back interviews — typically technical, behavioral, system design, and hiring manager — covering one candidate in a single day or block.

The candidate-led portion where the candidate questions the interviewer about the team, role, and company. A neglected reverse interview is one of the top reasons strong candidates decline offers post-loop.

A late-stage interview format where the candidate spends a half-day with the team they would join — meets their would-be manager, sees the codebase or roadmap, and gets unfiltered Q&A access. Trades pipeline speed for offer-acceptance rate.

Answer framework for behavioral questions: Situation, Task, Action, Result. Recommended by Amazon’s Leadership Principles guide and most structured-interview research as a way to surface specific, verifiable evidence rather than generalities.

A format where every candidate is asked the same predefined questions and rated against a fixed rubric. Meta-analyses (Schmidt & Hunter 1998; Sackett et al. 2022) consistently rank structured interviews among the strongest predictors of job performance.

An open-ended interview where the candidate sketches the architecture of a large-scale system — load balancing, data partitioning, caching, queueing, consistency trade-offs. Standard at senior+ engineering levels at FAANG and most tech companies.

A timed problem candidates solve outside the live interview. Trades scheduling flexibility for the risk of unbounded effort; bounded timeboxes mitigate this.

An offline coding, design, or analysis exercise candidates complete on their own time before a live interview round. Better signal-to-noise than whiteboarding when scoped tightly (≤ 4 hours), but adds load that filters out candidates with caregiving responsibilities — many teams now cap effort or pay for time.

An IRS-required independent appraisal of a private company’s common stock fair market value, refreshed at least annually or after material events. Sets the strike price for new option grants; relying on a stale 409A creates personal tax penalties for grantees.

Variable cash compensation tied to performance, company results, or sign-on. Annual target bonus is a percent of base; payouts may be 0–200% of target.

A revised offer from either the hiring company (responding to a candidate’s ask) or the candidate’s current employer (responding to a resignation). Current-employer counters retain the candidate ~10% of the time within 12 months — most who accept leave anyway, often on worse terms.

A raise or promotion offered by a candidate’s current employer after they resign. Industry data shows 50-80% of accepted counter-offers leave within 12 months — typically used as a stop-gap rather than a long-term retention play.

A new equity grant given to existing employees — typically annually after the initial cliff — to keep total equity competitive as the initial grant vests. Refresh sizes are usually 25-50% of the original new-hire grant.

The point ~3 years into a 4-year initial equity grant where annual refresh grants become the dominant retention lever. Ask about refresh policy explicitly during negotiation — companies that pay top-of-band base but skip refreshes are setting up a cliff exit.

A US employee stock option type with favorable tax treatment if held one year past exercise and two years past grant. Subject to AMT at exercise; $100,000 annual vesting limit per IRS §422 above which options convert to NSOs.

A US stock option taxed as ordinary income at exercise on the spread between strike and fair market value. Simpler tax mechanics than ISOs and the default option type for contractors and post-$100k ISO grants.

For roles with variable pay (sales, partnerships, some BD), OTE is base salary plus the commission/bonus assuming 100% quota attainment. Negotiate base AND commission rate — base is the floor when quota slips.

A defined min–mid–max salary range per level and geography that bounds offers and raises. Used to enforce internal equity and to comply with pay-transparency laws (e.g., NYC Local Law 32, Colorado EPEWA, EU Pay Transparency Directive 2023/970).

Additional equity granted to existing employees, typically annually or at promotion, to offset cliff drop-off and keep total compensation competitive. Common at public tech companies; rare at early-stage startups before Series B.

Company shares granted to employees that vest over time. Taxable as ordinary income at vest; valuation depends on current share price.

A one-time cash payment paid at the start of employment, typically used to bridge a candidate’s lost equity vest from a previous job or to win competitive offers. Almost always comes with a clawback clause: leave within 12–24 months and you repay a pro-rated amount.

Base salary plus bonus, equity (vested per year), sign-on, and benefits monetized. Apples-to-apples comparison metric across offers.

A period (commonly 1 year) where no equity vests; departure before the cliff means zero equity. After the cliff, equity vests on a regular schedule.

Laws prohibiting employers from asking about criminal history on initial job applications. Adopted in 37 US states and 150+ cities as of 2024 (NELP tracker). Conviction inquiries may still occur later in the process where legally permitted.

A GDPR-mandated contract between a data controller (employer) and processor (ZhiYin, ATS, background-check vendor) specifying scope, retention, sub-processors, and breach notification. Required for any vendor handling EU candidates' personal data.

The 2018 EU regulation that governs how personal data about EU residents is collected, processed, and stored. Grants data subjects the right to access, rectify, erase, restrict, port, and object to processing. ZhiYin honors deletion requests within 30 days and logs every PII access.

Provision of the EU General Data Protection Regulation giving individuals the right not to be subject to solely automated decisions with legal or similarly significant effects — including hiring. Requires meaningful human review of AI-assisted candidate decisions and the right to contest them.

A US federal requirement to verify every new hire's identity and employment authorization within 3 business days of their first day. Mishandled I-9s expose the employer to penalties of up to ~$2,700 per violation per employee.

Contract binding parties to keep specified information confidential. Common during interview loops that involve proprietary problems or roadmap discussions.

A clause restricting an employee from joining competitors for a period after leaving. Enforceability varies sharply by jurisdiction; many places limit or void them.

A clause restricting an employee from joining a competitor or starting a competing business for a defined window after departure. Enforceability varies sharply by jurisdiction (e.g., effectively void in California; bounded by reasonableness elsewhere).

Contractual lead time an employee must serve between resignation and final day. Two weeks is the US norm; one to three months is common in the EU, UK, and APAC for senior roles. Material to start-date negotiations.

Laws requiring employers to disclose salary ranges in job postings. Active in NY, CA, CO, WA, IL, MD, and the EU under Directive 2023/970 (member states must transpose by June 2026). Range disclosure is now table stakes for cross-state hiring.

A defined initial employment window (commonly 30, 60, or 90 days) during which either party can terminate with reduced notice. Used to validate fit on both sides. Some jurisdictions cap maximum probation duration by statute.

A GDPR Article 17 right that lets EU residents demand erasure of their personal data when it’s no longer necessary, when consent is withdrawn, or when processing was unlawful. Candidates can trigger this through the privacy center; processing is queued and confirmed via email.

Legal authorization for a person to be employed in a given country. Employers must verify before hire; visa sponsorship may be required for international candidates.

Statistical testing of an AI hiring tool to detect disproportionate negative outcomes for protected groups. The "four-fifths rule" — any protected group selected at less than 80% of the highest-selected group rate — triggers a compliance review. ZhiYin runs this monthly.

A defined class of AI tools used to make or substantially assist employment decisions. NYC Local Law 144 requires bias audits and candidate notice for AEDT use.

When a neutral selection process produces materially worse outcomes for a protected group. The EEOC four-fifths rule flags selection rate ratios below 0.8 for review.

EU regulation classifying employment-related AI as high-risk, with requirements for risk management, bias monitoring, transparency, and human oversight.

Further training of a pretrained LLM on a narrow, task-specific dataset so it specializes in that domain (e.g. resume scoring, interview transcript analysis). More expensive than RAG, but yields lower-latency, lower-token-cost inference.

A layer that picks between LLM providers (gateway, cloud, local) per request, with failover. ZhiYin routes gateway → OpenRouter → Ollama.

A class of attacks where malicious text inside user input (e.g. a resume PDF, a job description) coerces an LLM into ignoring its system instructions. Mitigated by input sanitization, output validation, and treating LLM outputs as untrusted data.

An LLM pattern that retrieves relevant snippets from a knowledge base (resumes, job descriptions, company guides) at query time and stuffs them into the prompt, so the model answers from current ground-truth rather than its training data. Reduces hallucinations and lets the model cite sources.

A numeric vector representing semantic meaning of text. Used for similarity search — e.g., matching a candidate bio to job descriptions.

A client-supplied unique value (header or field) that lets a server safely deduplicate retried requests so the same operation isn’t executed twice. Standard practice for payments (Stripe), checkout, and any mutating endpoint that may be retried after a network blip.

An API endpoint where repeated identical requests have the same effect as a single request. ZhiYin mutations (apply, unlock, refund) accept an idempotency-key header so retries from flaky networks never double-charge or double-apply.

Data that identifies an individual (name, email, phone, address). On ZhiYin, candidate contact information is hidden until the employer pays an unlock fee, and every access is audit-logged.

A guardrail that caps how many requests a single client (IP or user) can make per time window. ZhiYin tiers limits per-endpoint: write mutations stricter than reads, AI inference stricter than CRUD — protects against scrapers, brute force, and runaway client retries.

An HTTP callback delivered by a third party when an event occurs (e.g., Stripe payment succeeded). ZhiYin dedupes webhook deliveries to keep state idempotent.

A cryptographic signature (typically HMAC-SHA256) the sender includes on every webhook, computed over the raw body using a shared secret. Receivers verify it before processing — defends against forged callbacks from Stripe, Resend, and other providers.