Prepare and collect data

As a preparer, you manage the files moving between the source dataset and the labelers. You use Prepare mode to split and join files. The administrator supplies the configuration that defines the columns; you do not need to edit it to distribute work.

Before you start

Have a tested configuration, the source CSV, a list of labelers, and a place to track which batch each person receives. Keep an unchanged copy of the source file.

Check	What to look for
Correct project	Use the same configuration version for splitting, labeling, and joining.
Column headers	Every declared input column is present; spelling and capitalization match.
Stable IDs	Ask the administrator to carry a record ID into the output so results can be matched to the source.
Representative sample	Open several records and check dates, lists, and unusual or empty cells.
Writable folder	Labelers need permission to write exports beside their input files.

Send work out

1. Load the configuration and choose Mode → Prepare.
2. Drop one source file into the window.
3. Confirm Split, choose the number of parts, and review the filenames and row counts.
4. Run the split. Send each labeler a part, the matching configuration, and the labeling instructions.
5. Record the part name, recipient, configuration version, and return status in your batch list.

Review the split preview before creating files. Existing files will not be overwritten. Select image to enlarge.

Use batches small enough to finish and return regularly. Split preserves record order; it does not randomize the dataset or create overlapping assignments for agreement checking.

Collect the results

Ask each labeler for both *-output.csv and *-remaining.csv, if present. Keep the received files together with their batch information.

1. Drop the output files into Prepare together and confirm Join outputs.
2. Resolve errors. Review duplicate warnings against your batch list before proceeding.
3. Save the joined output under a new filename.
4. Separately join remaining files, then split or reassign them for another labeling pass.

See the full split and join instructions →

Choose the next guide

Task	Guide
Use the split/join controls	Split and join files
Keep an assignment register	Manage batches
Resolve errors and missing submissions	Check returned work

Check that everything came back

For each assignment, finished records plus remaining records should account for the input records. Investigate missing batches, repeat submissions, and unexpected counts before using the combined dataset. A duplicate warning is a prompt to review the files; it does not decide which person’s judgment to keep.

Remaining files contain original source values, not the labeler’s unfinished answers. If somebody needs to continue their partial answers, have them resume their saved session on the original computer before exporting.

Resolve common handoff problems →