Troubleshooting

This section provides guidance for solving common problems and issues encountered when using the ConversionFlow library.

Contents:

• Frequently Asked Questions
  • How to interpret the results?
  • Is a non-negative prior justified?
  • Is the Half-Cauchy prior justified?
  • Should the downstream nodes have the biggest weights?
  • What do convergence warnings mean?
  • The problem with exit nodes (actual car purchase)
  • How would individual user-level data change the model?
  • What about the acyclicity assumption (the ‘A’ in DAG)?
  • How can I adapt the model for different countries?
  • What if the optimizer is not finding good solutions?
  • How much data is needed for reliable results?
  • Can I integrate this with our ad platforms directly?