Back to Home

Troubleshooting

Solutions to common issues you might encounter when using qualcode.ai.

File Upload Issues

File format not supported

qualcode.ai supports the following file formats:

  • CSV (.csv) - Comma-separated values
  • Excel (.xlsx, .xls) - Microsoft Excel spreadsheets

If your file isn't uploading:

  • Check the file extension matches one of the supported formats
  • Try exporting from your spreadsheet software as CSV
  • Ensure the file isn't corrupted by opening it locally first

Encoding errors or garbled text

If your data contains special characters, accents, or non-English text that appears garbled:

  • Save your CSV with UTF-8 encoding
  • In Excel: File > Save As > choose "CSV UTF-8 (Comma delimited)"
  • In Google Sheets: File > Download > Comma-separated values (.csv)

File too large

Maximum file size is 10 MB. If your file exceeds this:

  • Remove unnecessary columns before uploading
  • Split large datasets into multiple files
  • Consider compressing text columns that contain very long responses

Column not found

When selecting a column to code, make sure:

  • The column has a header row (first row contains the column name)
  • The column name doesn't contain special characters that might cause issues
  • The column actually contains text data (not formulas or references)

Low Kappa Scores

Understanding low scores

A low Cohen's Kappa score (below 0.40) indicates poor agreement between the AI raters. This can happen for several reasons:

Possible Cause Solution
Ambiguous categories Review your category definitions and make them more specific
Overlapping categories Ensure categories are mutually exclusive or use multi-label mode
Vague descriptions Add concrete examples to your category descriptions
Domain-specific language Add training data with examples from your specific domain
Too many categories Consider consolidating similar categories

Improving accuracy

  1. Review disagreements: Go through the reconciliation queue to see where raters disagree
  2. Look for patterns: Are disagreements concentrated in certain categories?
  3. Refine descriptions: Update category descriptions to address common confusion points
  4. Add training data: Include reconciled examples as training data for future runs
  5. Re-run: Run again with the improved guide and training data

Self-learning workflow: qualcode.ai is designed for iterative improvement. Reconciled disagreements become training data automatically, so each subsequent run is sharper than the last. Academic coding schemes are typically developed through multiple rounds — qualcode.ai makes each round faster and more precise.

Credit Issues

Insufficient credits

If you don't have enough credits to start a coding run:

  • Check your credit balance in the top navigation bar
  • Try using a lower model tier (Budget instead of Standard or Quality)
  • Run a smaller subset of your data first
  • Purchase additional credits from the Credits page

Credits not refunded for cancelled run

When you cancel a coding run:

  • Credits for unprocessed responses are automatically refunded
  • Credits for already-processed responses are not refunded (processing already occurred)
  • The refund appears in your balance immediately
  • Check your credit transaction history in Settings for details

Unexpected credit usage

Credit usage depends on several factors:

  • Model tier: Budget (1x), Standard (1.5x), Quality (3x)
  • Training data: Larger training datasets slightly increase costs
  • Response count: Unclassifiable responses (empty, gibberish) are not charged

Check before running: The cost estimate shown before starting a run accounts for all these factors. Review it to avoid surprises.

Coding Run Failures

Run stuck at "Processing"

If a run appears stuck:

  • Large datasets (1000+ responses) take longer to process
  • Check the progress percentage - if it's changing, the run is still active
  • Wait at least 5 minutes before assuming there's an issue
  • If truly stuck, refresh the page and check the run status again

Run failed with error

If a run fails completely:

  • Credits for unprocessed responses are automatically refunded
  • Check the error message for specific details
  • Common causes: invalid coding guide configuration, API rate limits
  • Try running again after a few minutes

Partial results

If a run completes but some responses show errors:

  • Individual response errors don't fail the entire run
  • Check which responses failed in the results view
  • Common causes: extremely long responses, unusual characters
  • You can re-run just the failed responses

Export Issues

Export file is empty

Check that:

  • The coding run completed successfully
  • There are classified responses (not all were unclassifiable)
  • You selected the correct export format

Missing columns in export

Export includes different columns depending on the format:

  • Full export: All columns including both rater outputs and confidence scores
  • Summary export: Just the final classification and agreement status
  • See Export Formats for complete details

Account Issues

Can't log in

  • Check that you're using the correct email address
  • Use the "Forgot password" link to reset your password
  • Clear your browser cache and cookies, then try again

Email verification not received

  • Check your spam/junk folder
  • Add support@qualcode.ai to your contacts
  • Request a new verification email from the Settings page

Getting More Help

If you're still experiencing issues:

  • Check the FAQ for additional answers
  • Contact support at support@qualcode.ai
  • Include details about the issue: what you were doing, any error messages, your browser/device