Skip to content
Quality Framework

Quality is the product, so we measure it like one

Five components make excellence repeatable: a reviewer hierarchy, a three-tier pipeline, quantified metrics, an independent audit program, and written acceptance criteria in every engagement.

Acceptance
≥95%
Agreement
κ ≥ 0.7
Peer review
100%
Escaped
≤5 / 1k
Three-tier reviewCoverage narrows, authority rises

96.4%

Acceptance rate

SLA floor is 95%

κ ≥ 0.7

Agreement gate

Cleared before production

Quality Framework

QA is not our process.It is our product.

Marketplaces deliver raw output and leave verification with the buyer. We deliver verified output with the evidence attached.

Review coverage

  • Tier 3. AuditHead of Quality

    Random 10% plus risk-weighted sample

  • Tier 2. Senior reviewSenior Reviewers

    25–35% sample plus all flagged items

  • Tier 1. Peer reviewReviewers

    100% of contributor output

  • ProductionContributors

    Self-check against rubric checklist before submit

MetricTargetMature
  • Acceptance rate

    Share of delivered items accepted by the client without rework.

    ≥95%≥98%
  • First-pass yield

    Share of contributor items passing Tier 1 unchanged.

    ≥85%≥90%
  • Inter-rater agreement

    Cohen’s κ / Krippendorff’s α on the double-rated sample.

    ≥0.7≥0.75
  • Escaped-defect rate

    Client-found defects per 1,000 delivered items.

    ≤5≤2
  • Gold-task accuracy

    Contributor score on blind, seeded known-answer tasks.

    ≥90%≥93%
  • Review accuracy

    Tier 1 decisions overturned at Tier 2 or Tier 3.

    ≤8%≤5%

Rule of the house

No batch leaves Newbieget Labs below the client’s acceptance bar. If internal QA predicts a miss, the batch is reworked before delivery. Schedule slips are negotiable. Quality misses are not.

Audit program

Independent of the delivery line.

Audits are run by the Head of Quality, not by the people who produced the work. Findings drive rubric updates and coaching plans rather than blame.

Blind gold seeding
Continuous
3–5% of every queue is known-answer work, indistinguishable from live tasks. Scores feed contributor dashboards automatically. No one is graded on a sample they can see coming.
Weekly calibration
Weekly
The whole pod rates the same ten items. Disagreements are argued to resolution and the rubric clarification is versioned and logged. Drift is corrected weekly, not at delivery.
Monthly program audit
Monthly
The Head of Quality re-scores a stratified sample per program, independent of the delivery line. Findings drive rubric updates and coaching plans.
Client quality report
Quarterly
Metric trendlines, defect taxonomy and corrective actions. Sent unprompted, whether or not the quarter went well.
Three-tier reviewCoverage narrows, authority rises

96.4%

Acceptance rate

SLA floor is 95%

κ ≥ 0.7

Agreement gate

Cleared before production

Structural ratios

Constants, not targets.

These ratios hold at every scale we operate. Headcount never grows faster than the QA system can absorb it.

1 : 6

Reviewer to contributors

Holds at every scale

1 : 15

Senior reviewer to contributors

Owns the QA gate

1 : 1

Project lead to client program

Owns the SLA

1.3×

Bench to deployed demand

Calibrated capacity in reserve

Contract language

Acceptance criteria. Standard SOW clause

A delivered item is accepted if it (a) satisfies every mandatory rubric dimension at or above the agreed threshold, (b) passes the client’s spot-check sample at ≥[95]% within [10] business days, and (c) contains no Category-A defects. Factual error, safety violation, format break. Batches below threshold are reworked once at Newbieget Labs’s cost within [5] business days.

Bracketed values are negotiated per engagement. The structure is not.