FAQ
FAQ
Frequently Asked Questions
Answers to the questions we hear most often before an order —
on cost, quality, delivery, and security.
How much does annotation outsourcing cost?
It is fully usage-based: Bbox (object detection) from $0.036/rectangle, keypoints from $0.021/point, segmentation from $0.152/region, 3D cuboids (LiDAR) from $0.121/unit, and LLM multi-step evaluation from $0.121/evaluation (all excluding tax). Setup and management fees are $0; you pay only for data creation. We provide an exact quote for free after reviewing your requirements.
Is there a minimum order?
There is no minimum-order requirement. For image annotation, we flexibly take on PoCs from around 50 to 100 items.
Can I try it for free?
Yes, a free trial is available. We annotate a sample of your real data (around 10 to 50 items) using the actual team, so you can check quality and communication before you commit.
Is a quote free, and how soon can I get one?
It is free. After hearing your data type, volume, and specifications, we provide an actual-cost-only quote. If requirements are set, we typically respond within a few business days.
Can I order even if the specification is not finalized?
No problem. Drawing on 1,000+ completed projects, we support you from the requirement-definition and annotation-spec design stage and document the judgment criteria.
What data types do you support?
Images and video (object detection / segmentation / pose estimation / tracking), 3D point clouds and LiDAR, LLM (RLHF/DPO, evaluation, SFT), text and speech, and data collection and preprocessing.
How is quality guaranteed?
Rather than sampling inspection, we perform full-volume acceptance inspection, maintaining 99.7% quality consistency. For one year after delivery, we fix at no charge any defects on our side that deviate from the agreed spec (this does not cover spec changes).
What are typical delivery times?
It varies by project, but setup is as fast as 2 days. Large projects can be parallelized with our global team; for example, we have delivered 100,000 items in 2 months.
Can you build Japanese RLHF / LLM-evaluation data?
Yes. Native Japanese annotators create preference data (RLHF/DPO), multi-step evaluations, and instruction-response pairs (SFT), and we support you from the evaluation-criteria design stage.
What is the flow from order to delivery?
1) Inquiry & requirement hearing → 2) Spec design & quote → 3) Free trial (optional) → 4) Order & annotation → 5) Full-volume inspection & delivery.
Tell me about your security.
Under a management framework based on ISO/IEC 27001 (ISMS), we enforce no-retention, no-local operation, sign NDAs with 100% of staff, and separate teams per project. We never repurpose your data to train our own AI. See our Security page for details.
What are the contract terms?
We work primarily under a service (outsourcing) agreement. We also handle individual NDAs and responses to security checklists, so please feel free to reach out.
Still have questions? Talk to us.
Even if your requirements aren't finalized, that's fine.
Our specialists will suggest the best approach.
