> For the complete documentation index, see [llms.txt](https://docs.unitlab.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.unitlab.ai/documentation/get-started/production-readiness-review.md).

# Pilot checklist

Test your setup on a small sample before you label your whole dataset.

Before you assign thousands of files, run a small pilot: 10 to 50 typical files plus a few hard ones. A pilot finds unclear rules, missing classes and export problems while they are still cheap to fix.

## 1. Set up

* [ ] The files you will label are uploaded and open correctly in the Workbench.
* [ ] Related files (for example several camera views) are [grouped](https://docs.unitlab.ai/documentation/data/data-groups-and-layouts.md) into one task.
* [ ] The [ontology](https://docs.unitlab.ai/documentation/ontologies/ontologies-overview.md) has every class and property you need, with the right shape (box, polygon, mask...).
* [ ] [Instructions](https://docs.unitlab.ai/documentation/projects/project-instructions.md) include at least one good example and one bad example per class.
* [ ] The [workflow](https://docs.unitlab.ai/documentation/workflows/workflows-overview.md) has a Review stage, and **Rejected** goes back to annotation.

## 2. Label and review

* [ ] An annotator labels the pilot files using only the instructions.
* [ ] A reviewer rejects at least one file, and you confirm it comes back to the annotator with the comment.
* [ ] Hard cases (blur, occlusion, tiny objects) are discussed and the instructions are updated.
* [ ] If you use AI-assisted labeling or a Model stage, its output is checked by a person.

## 3. Export and check

* [ ] You [create a release](https://docs.unitlab.ai/documentation/releases/create-a-release.md) from the reviewed pilot files.
* [ ] You open the export in your training tool and the classes, shapes and counts are correct.
* [ ] Polygons with holes, masks, keypoints and properties survive the format you chose.

## 4. Decide

If every box is checked, assign the rest of the data. If not, fix the setup and repeat the pilot on new files.

{% hint style="info" %}
Repeat this checklist whenever you change the ontology, workflow, data type or model.
{% endhint %}

## Related

[Quickstart](https://docs.unitlab.ai/documentation/get-started/end-to-end-quickstart.md) · [Review and rework](https://docs.unitlab.ai/documentation/workflows/review-rework-and-escalation.md) · [Download and validate a release](https://docs.unitlab.ai/documentation/releases/inspect-download-and-validate.md)
