Optimal Floors Discovery¶
Dynamic Floors leverages machine learning models to dynamically optimize incoming Supplier requests by adjusting bid floors.
Bid Floor Impact¶
Increasing the bid floor can lead to some Buyers bidding more aggressively, potentially improving competition and revenue.
Main Objectives¶
Maximize revenue opportunities: balance dynamic bid floor adjustments for Open Auction to achieve higher revenues while maintaining auction efficiency.
Ensure flexibility and control: allow admins to enable dynamic floors on specific trading pairs.
Traffic groups¶
Incoming traffic is split into three groups so the models can be trained and measured without affecting all inventory.
Group |
Description |
|---|---|
holdout |
Traffic LiteSwitch never touches. Baseline for measuring success. |
learning |
A random factor is applied to the request floor. Produces training data for the models. |
optimal |
Floor chosen by the models. Production behaviour. |