How would you determine the pay structure for data labeling teams?

To establish a pay structure for data labeling teams, I would first clarify the scope of work, including whether it's in-house or outsourced, the data modality (text, images, etc.), and the complexity of the labeling tasks (e.g., bounding boxes, segmentation). I'd also assess any historical quality issues.

My primary goal would be to incentivize high-quality output, recognizing that poor-quality labels directly impact downstream model performance. Therefore, the pay structure should reflect this critical need for accuracy and consistency.