SOLUTIONS · EXPERTINI ATS

How Expertini ATS Ranks Applicants

The mechanics of turning a stack of CVs into an ordered, explainable shortlist.

3 min read · Updated July 2026 · Expertini Editorial

This page walks through what actually happens between a candidate clicking "Apply" and a recruiter seeing an ordered, scored shortlist — the practical mechanics behind the Candidate Match Score.

▶ Watch the 30-second overview — no sign-up needed

Step 1 — PII stripping
Step 2 — dimension extraction
Step 3 — deterministic scoring
Step 4 — the ranked shortlist
What the recruiter actually sees on each application
What ranking deliberately does not do
Why this ordering is defensible later
How this is architected in the platform
Operational and audit posture
Frequently asked questions

01Step 1 — PII stripping

The uploaded CV's text is run through pattern-based stripping that removes names, emails, phone numbers, addresses, and demographic-suggestive language before anything else happens to it.

02Step 2 — dimension extraction

The stripped CV and the job description go to Gemini, which identifies 5-9 competency dimensions specific to that job posting (not a generic fixed rubric) and, for each, whether the CV shows evidence of it — with a short citation of that evidence.

03Step 3 — deterministic scoring

Ordinary code — not the AI — computes CMS = Σ(CSS×JRIS)/Σ(JRIS) from that extraction. Any dimension marked strictly mandatory in the job description that shows zero evidence in the CV is pinned to zero as a hard blocker, regardless of how strong the rest of the CV is.

04Step 4 — the ranked shortlist

Applicants land in the pipeline with their score, tier (Exceptional through Low), and dimension-level rationale visible to the recruiter — who still makes the actual hiring call. Nothing here auto-rejects a candidate; hard blockers are flagged prominently, not silently filtered.

05What the recruiter actually sees on each application

Opening a scored application shows three layers, in order of how often they get used. The score and tier band answer "roughly where does this person sit" at a glance while triaging fifty applicants. The dimension list answers "why" — each of the 5-9 competencies extracted from the job description, with the candidate's strength on that dimension and a short citation of the CV evidence behind it. And the hard-blocker flags answer "is there a disqualifier I need to look at before anything else." The design assumption is that a recruiter will read the number for most candidates and the rationale for the borderline ones — which is exactly where an explanation earns its keep.

06What ranking deliberately does not do

Three things are absent by design, not by omission. There is no automatic rejection threshold — the platform will never silently discard an application below some score, because a cutoff that makes sense for one role's applicant pool is wrong for another's, and that judgement belongs to the person who can see the whole pool. There is no cross-candidate normalisation — a candidate's score depends only on their own CV and the job description, never on who else happened to apply that week, so scores stay comparable across time. And there is no personality, culture-fit, or video-analysis inference — categories of automated assessment with weak evidence bases and serious fairness problems that we've chosen to stay out of entirely. The fuller reasoning behind these boundaries is on the AI Ethics Guide.

07Why this ordering is defensible later

Because every step is either deterministic code or an extraction with cited evidence, a ranking produced today can be explained in eighteen months — to a hiring manager who asks why someone was shortlisted, to a rejected candidate exercising a right to review, or to an auditor asking how the tool works. The Candidate Match Score methodology is published in full, so "how does your ATS rank applicants" has a checkable answer rather than a trust-us one. If you're comparing platforms, ask each vendor to walk you through these same four steps for their product and note where the answer becomes vague.

Engineering notes

Platform architecture & operations

A1How this is architected in the platform

How Expertini ATS Ranks Applicants is not a bundle of point products — it is a slice through one platform. The platform is deliberately server-rendered: every view is prepared by the application server and shipped as complete HTML, with no client-side framework, no third-party CDN scripts, and no build pipeline between the data and the page. What renders is what the server computed — the property that makes the interface auditable.

All persistence runs on a single search-native document store; every query carries the organisation's identifier as a mandatory filter at the lowest query layer. Tenant isolation is therefore structural — a property of how every request is composed — rather than a policy that relies on application code remembering to check.

Every capability referenced on this page resolves to a registered tool or connector: the tools directory and the integrations catalogue are renderings of the same registries the application enforces at runtime, so what this page describes and what the product gates can never drift apart.

A2Operational and audit posture

Screening is deterministic and published — the same inputs produce the same outputs, hard requirements block rather than average away, and the methodology is public on the research page. Actions that touch external systems are explicit and journalled per event; usage reporting aggregates the same journals the actions write, not a parallel telemetry system.

Anything that leaves the request path — notification fan-out, webhook delivery, activity journalling, mail — runs in fire-and-forget background threads. A slow external endpoint can never make the interface hang, and a failed side effect is logged rather than silently retried into inconsistency.

Everything written is yours to take: CSV exports and the Data Export app cover the same stores the product itself reads. The exit is as open as the entrance — by design, not concession.

Frequently asked questions

Can I set a score below which applicants are rejected automatically?⌄

No — and that's deliberate. Scores rank and inform; rejection is a human action. Hard blockers give you a prominent flag for missing mandatory requirements, which most teams use as their practical filter, but the decision to act on it stays with the recruiter.

Does a candidate's score change depending on who else applied?⌄

No. Each score is computed from that candidate's CV and the job description alone — there's no curve and no cross-candidate normalisation, so a 74 scored in March means the same thing as a 74 scored in September.

Can candidates see their own score?⌄

The score and rationale are recruiter-facing pipeline tools. What a candidate experiences is described on the Candidate Experience page — including their right to request human review of any decision an automated score influenced.

How long does scoring take after someone applies?⌄

Scoring runs when the recruiter triggers it against a job's applicants, subject to the monthly CMS budget on your plan — see pricing for the per-plan limits. It's a deliberate action, not a silent background process burning your quota.

At a glance

PII stripped before the CV reaches any AI step
5-9 dimensions derived fresh per job, not a fixed rubric
Score computed by code, not by asking the AI for a number
Hard blockers are flagged, not silently auto-rejected
No hidden auto-reject threshold, ever
Scores independent of who else applied — comparable across time

See how expertini ats ranks applicants on your own hiring.

Bring a real job description to a 30-minute demo — free trial included.

Book a demo