Engagement models

Four ways to buy, one delivery standard

Every model carries the same QA pipeline, the same acceptance criteria and the same evidence pack. What changes is how the work is scoped and billed.

Get a quote Quality framework

Four ways to buyOne delivery standard behind all of them

Same QA, same SLA

01Most common

Managed project delivery

A defined dataset or evaluation with a known shape

Fixed scope and fixed acceptance criteria. We staff a named pod, calibrate it against your rubric, produce, review and deliver with the evidence attached.

Billing: Fixed-scope or per-unit
Commitment: Per project

What it includes

Named pod with a project lead who owns the SLA
Calibration gate cleared before production starts
Three-tier QA with 100% first-pass review
QA evidence pack with every batch

The default entry point, and how most pilots convert.

Dedicated capacity pod

Continuous throughput where volume is predictable

A named team reserved for you and embedded in your workflows, typically contributors plus reviewers plus a lead. The pod holds its calibration between batches rather than re-ramping.

Billing: Monthly retainer
Commitment: Three-month initial term

What it includes

Reserved, named team that stays calibrated on your rubric
Volume commitment with a quality SLA
Optional fee-at-risk against the acceptance bar
Bench cover so absence does not stall throughput

Once a pilot has converted and the work is ongoing.

QA as a service

Verifying output you already have from another source

We operate a review and audit layer over work produced by your own crowd or another vendor. Findings go to you unedited, with the evidence trail.

Billing: Per item or retainer
Commitment: Rolling

What it includes

Defect taxonomy and acceptance definition agreed up front
Stratified, risk-weighted sampling
Per-batch metrics and monthly findings
Independent of any Newbieget Labs delivery line

The lowest-risk way to start, because nothing depends on us producing first.

Benchmark engineering

Building an evaluation harness rather than filling one

Task taxonomy, environment engineering, verified gold solutions, grading pipeline and reporting, run as one accountable workstream and handed over complete.

Billing: Project pricing
Commitment: Per program

What it includes

Dockerised, reproducible task environments
Gold solution verified from a clean container
Adversarial review for shortcuts and leakage
Harness handed over to run in your CI, not ours

When the evaluation itself is the deliverable.

Pricing

Why there is no price list

Rates move with domain, expertise tier, verification depth and volume. A clinical-safety review hour and a general-tier ranking hour are not the same unit, and publishing a single number would misprice both. Send the scope and you get a real quote, not a range.

Send a scope Size it first

Quotes are fixed for the scope described, not indicative
Pilot pricing is quoted before any commitment to ongoing work
Rework against a missed acceptance bar is not billed
No minimum retainer to run a first pilot

What we carry

Commitments with a cost attached.

A quality claim only means something if missing it is expensive for the vendor rather than the buyer.

Acceptance-rate SLA

A written acceptance threshold per engagement, with a fee-at-risk option where you want us to carry commercial exposure to it.

Rework at our cost

If a delivered batch falls below the agreed bar, it is reworked within the SOW window and you are not billed for the second pass.

Evidence, not assertion

Every delivery ships with the review trail, agreement data and gold-task results, so acceptance is a check rather than a matter of trust.

No lock-in

Work happens inside your tooling, rubrics are yours, and benchmark harnesses are handed over to run in your infrastructure.