A guide to learn about the variant testing process in Seatext AI

Seatext AI employs a custom GPT-4 model to generate multiple text variants.

This model is continuously trained on the results of Multiple Variants Testing (MVT) to improve its performance.

Instead of traditional A/B testing, which compares two variants, Seatext AI uses MVT. This method involves testing hundreds of text variants simultaneously.

MVT employs an enforced vectors technique, a machine learning method that assumes results from a smaller subset of the test as accurate and applies them across the entire test.

This technique allows Seatext AI to quickly identify and eliminate underperforming variants. Variants are discarded when they perform below a 65% confidence score.

The technique enhances the model’s efficiency, allowing it to deliver results up to five times faster than traditional A/B testing.

Seatext AI continuously monitors the performance of each text variant. Performance metrics include conversion rates and confidence scores for each variant.

Successful variants, those that show the highest engagement and conversion rates, are automatically implemented on the website.
The system updates the website with the best-performing text variants, ensuring ongoing optimization without manual intervention.

Users can view an overview of the testing results through the AI main hub in the Seatext interface.
For detailed analysis, the "variant edit" section provides insights into the performance of each text variant, including estimated conversion rates and confidence scores.

The insights gained from one client’s testing can benefit other clients. Seatext AI leverages this shared data to continuously refine its model, improving text generation and optimization across its user base.