Methodology

How the Human-First Score is calculated. Open, reproducible, no pay-to-rank.

Why a score at all?

Candidates want a simple signal: will this employer treat my application like a human being or like a row in a spreadsheet? The Human-First Score is a single 0–100 number derived from structured fields only — never from free-text reviews, sponsorships, or editorial opinion.

Inputs

Every report collects only four structured signals plus an optional anonymous comment:

  • Outcome — automated rejection, human response, no response, or interview.
  • Response time — within 24h, within a week, within a month, over a month, or never.
  • Rejection email received — yes / no.
  • Country — for cross-EU comparisons.

The formula

For each report we assign an outcome weight and a response-time weight, then average per company:

outcome_weight:
  interview            = 100
  human_response       =  80
  automated_rejection  =  40
  no_response          =  10

response_weight:
  within_24h    = 100
  within_week   =  80
  within_month  =  50
  over_month    =  20
  never         =   0

report_score  = 0.7 * outcome_weight + 0.3 * response_weight
company_score = mean(report_score for company)

Companies with fewer than 3 reports are shown as "not enough data" and are excluded from rankings to prevent single-candidate bias.

What we deliberately do NOT do

  • We do not accept money to improve, hide, or remove a score.
  • We do not run sentiment analysis on free-text comments.
  • We do not use the Human-First Score itself as a hiring signal — it is a compliance-and-trust signal, not a candidate-quality signal.
  • We do not publish individual reporters. Everything is anonymised on publication.

Abuse resistance

  • Per-IP rate limiting on submissions.
  • Honeypot fields and server-side schema validation.
  • Outlier reports (e.g. a single flood from one source) are weighted down.
  • Companies can post a factual response to any report.

Open dataset

Aggregated, anonymised report data will be released under an open-data licence once the dataset is large enough to be useful without risking re-identification. Researchers and journalists are welcome to contact us for early access.

Changes

Changes to the formula will be announced publicly and historical scores will be re-computed so rankings remain comparable over time.