Optimal Exploration of New Products under Assortment Decisions
How platforms like Amazon should balance new product exploration with revenue, according to new research.
Researchers from MIT and Cornell have published a paper on arXiv (2604.18800) titled "Optimal Exploration of New Products under Assortment Decisions," addressing how platforms like Amazon or Netflix should learn about new products when making assortment decisions. The study focuses on social learning: when customers purchase a new product and leave a review, its quality is revealed to both the platform and future customers. Since reviews require purchases, the platform must feature new products in the assortment ("explore") to generate reviews, but this exploration is costly because customer demand for new products is lower than for incumbent products.
The paper answers two key questions. First, should the platform offer a new product alone or alongside incumbent products? The authors show it is always optimal to pair the new product with the top incumbent products, despite the lower short-term revenue. Second, with multiple new products, should the platform explore them simultaneously or one at a time? The optimal number of new products to explore simultaneously increases with the "potential" of the new products and, surprisingly, does not depend on their individual purchase probabilities. The paper also finds that two canonical bandit algorithms, UCB and Thompson Sampling, both fail in this setting for opposite reasons: UCB over-explores while Thompson Sampling under-explores.
- Pairing new products with top incumbents is always optimal for balancing exploration and revenue.
- Optimal number of new products to explore simultaneously depends only on their potential, not individual purchase probabilities.
- UCB over-explores and Thompson Sampling under-explores in this capacity-constrained setting.
Why It Matters
Provides actionable insights for platforms to optimize new product launches and maximize long-term revenue.