New-Listing Exposure and Diversity Without Killing Conversion

After launching a strong booking-probability ranker, two problems appear in dashboards: (a) listings created in the last 30 days receive under 1% of impressions although they are 6% of supply, and (b) for city-centre searches page one is often 20 near-identical studios from the same property manager.

1. Explain why a well-trained ranker produces both effects.
2. Design the mechanism for each, including how you would bound the cost to guest conversion.
3. Which metrics would you add to the A/B scorecard to decide whether the changes are net positive?