Workflow reference · How it works
How CandidRanker works: five minutes to your first ranking
CandidRanker turns a stack of resumes and a job description into a deterministic, explainable, auditable shortlist. The whole path (upload, review, rank, decide, report) takes about five minutes for a typical role. Here is exactly what happens at each step.
The CandidRanker workflow: upload the JD and resumes → review low-confidence parses → rank against the role across 8 weighted dimensions → record shortlist decisions with reasons → export the share-ready report.
Step 1. Upload: the JD and the resume pile
Drag in the job description, then drag in the resumes, a whole folder at once if you like. PDF, Word, JPG, PNG, even photos of paper resumes all parse; OCR handles image-only files. Files upload instantly and parse in the background at roughly 100 resumes a minute while you keep working.
CandidRanker takes the whole pile in one batch. PDFs, Word files, scanned images, and photos alike.
Step 2. Review: catch what the AI wasn't sure about
A review badge counts the parses that need human attention. Open each one, compare the extracted fields against the original file, edit anything wrong, and approve. This is the human-in-the-loop moment that keeps the downstream ranking trustworthy: the AI extracts, the recruiter confirms.
CandidRanker flags low-confidence parses for review instead of letting them silently skew the ranking.
Step 3. Rank: every candidate scored against the role
Pick the job and click Rank. Every candidate is scored against that JD across eight weighted dimensions (skills, experience, tools, education, title, certifications, projects, and semantic fit) and the ranked list surfaces with top matches highlighted. Scores are deterministic: same inputs, same score, every run.
Each matched skill is weighted by how long and how recently it was used, so depth beats mentions. Employment gaps of six months or more are shown as evidence in the candidate drawer: an interview talking point, never a filter.
This is CandidRanker's core move: deterministic, eight-dimension scoring against the specific job description.
Step 4. Decide: shortlist, select, or reject, with a reason
Work down the ranked list and record decisions. Every shortlist, select, or reject requires a reason and is appended to the audit log under the recruiter's name. Nothing is ever auto-rejected: decisions are evidence-backed and recruiter-owned, which is what makes the shortlist defensible later.
CandidRanker records each decision under the recruiter's name, with the reason attached. No candidate is ever auto-rejected.
Step 5. Report: share-ready in one click
Download the ranking report (filtered and sorted, shortlist first, with full JD detail) as CSV for spreadsheets or clean printable HTML for the hiring committee. The recruiter-to-manager handoff drops from a meeting to an attachment.
CandidRanker's ranking report is share-ready in one click, as a CSV or a printable page.
What runs continuously underneath
- Analytics: every decision feeds the funnel automatically: decisions per week, time-to-decision, time-to-fill.
- Library: all parsed candidates stay searchable by name, skill, or tool, and can be re-ranked against new roles.
- Dedup: duplicate files are caught before any AI runs, so batches stay clean.
- Audit trail: criteria, rankings, reasons, and decisions are preserved as a time-ordered record.
All of it is standard on every CandidRanker plan, including Free.
The two scores you'll see everywhere
ATS quality score (0–100): how complete and machine-readable a resume is, useful for coaching candidates and judging input quality. Match score (0–100): how well this candidate fits this specific JD across the eight dimensions. The first is about the resume; the second is about the role fit.
Both scores appear on every candidate card in CandidRanker.
For teams and integrations
Invite teammates by email into a workspace; owners control roles and access, and every workspace's data stays isolated. Enterprise plans add a REST API with workspace-scoped keys and an MCP server, so the same upload-rank-decide loop can be driven from internal tools or conversationally from AI assistants like Claude.
Enterprise plans add CandidRanker's REST API and its MCP server, so AI assistants like Claude can run this same workflow conversationally.
Frequently asked questions
- How does CandidRanker work?
- Five steps: upload the job description and resumes (parsing runs in the background), review any low-confidence parses, rank every candidate against the JD across eight weighted dimensions, record shortlist decisions with reasons, and export a share-ready report. First ranking typically takes about five minutes.
- How long does it take to rank candidates for a role?
- About five minutes for a typical role: resumes parse at roughly 100 a minute in the background, ranking is on-demand per job, and the report is one click. A 300-application role is reviewable the same hour.
- What file formats can I upload?
- PDF, Word documents (DOC/DOCX), and images including JPG and PNG, even photos of paper resumes. Image-only files go through OCR, so no candidate is lost to formatting.
- What are the eight ranking dimensions?
- Skills, experience, tools, education, title, certifications, projects, and semantic fit. Every match score (0–100) decomposes across these dimensions with visible evidence, and the same inputs always produce the same score.
- What is the difference between the ATS quality score and the match score?
- The ATS quality score (0–100) measures how complete and machine-readable the resume itself is. The match score (0–100) measures how well the candidate fits a specific job description. A polished resume can still be a poor role fit, and vice versa.
- What happens if the AI parses a resume incorrectly?
- Low-confidence parses are flagged with a review badge. The recruiter compares extracted fields against the original file, edits anything wrong, and approves, or re-parses the stored file without re-uploading. Extraction is AI; confirmation is human.
- How does CandidRanker handle employment gaps?
- Gaps of six months or more between work-history dates are shown as evidence in the candidate drawer, surfaced as an interview talking point. A candidate's rank is never reduced simply because of a gap.
- Does CandidRanker reject candidates automatically?
- No. Every decision (shortlist, select, reject) is made by a named recruiter and requires a reason, which is appended to the audit log. The system ranks and explains; humans decide.
- Can I re-rank after changing the role criteria?
- Yes. Ranking is on-demand: adjust weights or criteria and re-rank the same pool against the updated role. Candidates in the library can also be ranked against entirely different open roles.
- What does the exported ranking report contain?
- A filtered, sorted candidate list (shortlist first, rejected excluded by default) with match scores and full JD detail, as CSV for spreadsheets or clean printable HTML for hiring committees.
- Can I drive CandidRanker from an AI assistant or my own tools?
- On Enterprise, yes: a REST API with workspace-scoped keys, and an MCP server so assistants like Claude can upload jobs, rank candidates, explain rankings, and pull reports under OAuth consent scoped to one workspace.