Uncharted Waters: Selling a New Product Robustly

Technical Summary: Uncharted Waters: Selling a New Product Robustly

Problem Statement
The paper addresses the strategic interaction between a seller of a new product and a buyer when both parties face uncertainty regarding product fit and the buyer's outside options. The seller sets a price and chooses an information provision policy (e.g., trials, samples) before the buyer decides whether to purchase immediately or incur a search cost to evaluate an outside option.

The core challenge lies in the seller's information asymmetry regarding the buyer's environment. While the buyer knows the distribution of her outside option, the seller only knows its mean ($\xi$) and support bounds. The seller cannot specify a precise probabilistic model of the buyer's search environment due to the novelty of the product and short demand histories. Consequently, the seller seeks a robust selling strategy—a price and information policy that maximizes the guaranteed revenue across all outside-option distributions consistent with the known mean and bounds.

Methodology
The author models the interaction as a two-stage game involving a monopolist seller, a buyer, and an adversarial "Nature":

1. Seller's Choice: The seller selects a price $p$ and an information provision policy, represented as a distribution over posteriors $H$ (where the expected posterior equals the prior probability of a high match value, $\mu$).
2. Nature's Choice: An adversarial decision-maker selects an outside-option distribution $G$ from the set of distributions with mean $\xi$ to minimize the seller's expected revenue.
3. Buyer's Decision: The buyer observes $p$ and a signal realization (posterior $w$). She compares the net value of the seller's product ($w-p$) against the expected benefit of search. If the net value exceeds a reservation threshold determined by $G$ and the search cost $s$, she buys immediately ("safe demand"). Otherwise, she pays $s$ to observe the outside option's realization and buys from the seller only if the outside option is worse.

The seller's objective is to solve a max-min problem:
$$\max_{(p, H)} \inf_{G \in M(\xi)} \Pi(p, H \mid G)$$
where $\Pi$ is the expected revenue. The analysis utilizes the properties of affine segments in the distribution of posteriors to hedge against distributional uncertainty and characterizes the trade-off between generating "safe demand" (buying without search) and "return-after-search demand."

Key Contributions and Results

1. Dual Roles of Information Provision
The paper identifies two distinct functions of information provision under robustness concerns:

• Hedging: By using an information policy that creates an affine segment in the distribution of posteriors (specifically, "uniform information"), the seller makes demand depend only on the known mean of the outside option ($\xi - s$) rather than its unknown shape. This hedges against the adversarial choice of $G$.
• Search Deterrence: Information can induce "safe demand" by generating sufficiently favorable posteriors ($w$) such that $w - p \geq \max_G (\text{reservation value})$. This deters search regardless of the specific outside-option distribution.

2. The Trade-off with Price
Price governs the relative value of these two roles. A higher price increases the margin per sale but makes it harder to create safe demand (as the buyer's net value $w-p$ decreases).

• Low Search Costs: When search costs are low, creating safe demand requires a price so low that it is suboptimal. The seller instead charges a higher price ($p^* > s/\xi$) and relies on uniform information (noisy, continuous distribution of impressions) to hedge demand.
• High Search Costs: When search costs are high, the price required to create safe demand ($p^{**} = s/\xi$) is sufficiently high to be profitable. The seller switches to full information (revealing the match value perfectly) to maximize the probability of safe demand.
• Intermediate Search Costs: The optimal strategy depends on the prior probability of a good match ($\mu$) and the mean outside option ($\xi$).
  • If the product looks unpromising ($\mu$ is low), the seller uses uniform information.
  • If the product looks promising ($\mu$ is high), the seller may use mixture information (a combination of a mass point at the highest posterior to create safe demand and a uniform segment for hedging) or full information, depending on the attractiveness of the outside option.

3. Comparative Statics and Counter-Intuitive Findings

• Non-Monotonic Pricing: The robust price is not monotone in search costs. As search costs rise, the price initially increases. However, at a critical threshold, the seller switches from a hedging strategy (high price, noisy info) to a deterrence strategy (lower price, informative info). This switch causes a downward jump in the optimal price.
• Information Quality: Generally, higher search costs lead to more informative product-fit information. Lower search costs induce the seller to provide noisier, less decisive information to maintain a high price.
• Buyer Welfare: Contrary to the presumption that easier search benefits consumers, the paper demonstrates that lower search costs can reduce buyer welfare. When search costs decrease, the seller may switch to a strategy with a higher price and noisier information, leaving the buyer worse off than if search costs were slightly higher.

Significance
The paper provides a theoretical framework for understanding why new products exhibit diverse pricing and information strategies. It explains why revolutionary products (high $\mu$) might receive full disclosure, while evolutionary or trade-off products might receive noisy or partial information.

Crucially, the results challenge the standard welfare intuition regarding search costs. In markets characterized by seller uncertainty about buyer alternatives, technological advances that reduce search costs do not necessarily benefit consumers; they can incentivize sellers to raise prices and obscure product-fit information. The study highlights that robustness concerns and search frictions jointly shape the optimal design of price and information, a dynamic that is absent in models where the seller knows the full distribution of outside options or where search is frictionless.