Glossary
What Is a Data Contributor in Robotics Data Collection?
Last reviewed: September 2026
Short answer
A data contributor is a screened, consented and paid person who performs tasks while wearing a capture rig, against a written skill spec, to produce robotics training data. Contributors are the source of every egocentric episode, so their screening, consent and adherence to the spec determine the dataset’s quality and licensability.
01
Why it matters
A dataset inherits the properties of the people who produced it. Contributors who drift from the spec lower the share of hours that pass QA; missing consent makes footage unusable in a commercial model; a narrow contributor pool limits diversity. A well-run contributor workforce is what makes custom collection reliable at scale.
02
How Firsthand handles it
Firsthand works with a vetted global crowd across 150+ countries and 500+ languages and locales. Contributors are screened and paid through the same QA and consent flow every time, a consent chain is recorded for each one, and only episodes that pass QA count as validated hours.
03
What is recorded for every contributor
| Record | Why it matters |
|---|---|
| Screening | Consistent quality across many people |
| Consent chain | Proof of agreement for commercial use |
| Spec adherence | Episodes that are comparable across locations |
| Validated output | Only QA-passing hours count toward delivery |
05
Frequently asked questions
- Are data contributors paid?
- Yes. Firsthand contributors are screened and paid through the same QA and consent flow every time.
- How are bystanders handled?
- Faces and license plates are anonymized before delivery, with human review on top of automated detection.
- Where are contributors located?
- Across 150+ countries and 500+ languages and locales.