# Promotional Recommendations  
Improved retail promotional targeting using multi-task recommendation models.

## Key Takeaways:

#### Promotional content lacked personalization  
Inconsistent metadata and fragmented content limited the effectiveness of promotional recommendations.

#### We upgraded recommendation infrastructure  
A new model was deployed to personalize promotions across channels and objectives.

#### Multi-task recommendation modeling  
A Two-Tower TensorFlow model was implemented to jointly optimize click-through and conversion signals using both long-term and short-term user behavior.

#### Significant gains in recommendation accuracy  
Improved ranking quality translated into more effective promotional targeting.