Introduction
Comparison Shopping Agents (CSAs) such as PriceGrabber or Bizrate generate revenue through referral fees, earned when users click through to a seller's website. However, the low barrier to switching between agents means users frequently query multiple CSAs to ensure they have found the absolute lowest price. This "search leakage" represents a direct loss of revenue for the initial agent.
While traditional solutions focus on technical improvements—such as increasing search speed or crawling more retailers—this paper investigates a psychological approach. The authors explore whether a CSA can influence a user's belief about the market to encourage them to stop searching, even without discovering new, lower prices.
The Belief-Adjustment Method
The core contribution of this research is the Belief-Adjustment Method. This approach moves away from the standard "bulk" display, where all prices are shown at once, in favor of a sequential presentation designed to exploit known cognitive biases. The method organizes price data into three distinct phases:
- Phase I: Anchor: The agent first presents a set of prices that are slightly higher than the absolute minimum it has found. This establishes a "reference point" or anchor in the user's mind. Crucially, the best price is withheld during this initial phase to set expectations.
- Phase II: Effort: The agent then reveals the "best" price. By presenting it after a series of higher prices, the system leverages the "effort heuristic"—the psychological tendency to perceive items found after significant search or work as more valuable or rare.
- Phase III: Despair: Finally, the agent shows the remaining prices, all of which are significantly higher than the best price. This "despair" phase is intended to convince the user that the market is exhausted and that further searching on other sites would be a waste of time.
Experimental Setup
The researchers evaluated their method using real-world price data for five distinct products (a keyboard/mouse combo, GPS navigator, ink cartridge, flash drive, and laser printer) sampled from major CSAs like Amazon and PriceGrabber.
To test the effectiveness of the ordering, they recruited 266 participants via Amazon Mechanical Turk to interact with an emulated CSA. The study compared the Belief-Adjustment Method against two baselines:
- Bulk Display: Showing all prices simultaneously (the industry standard).
- Random-Sequential: Showing prices one by one in a random order.
Key Results and Insights
The study found that the specific psychological ordering (Anchor -> Effort -> Despair) was the critical factor in changing user behavior. Simply switching to a sequential display (the random-sequential baseline) yielded no significant improvement over the bulk method.
The Belief-Adjustment Method outperformed the standard bulk method across every product tested, achieving an average 78.32% improvement in termination probability. For the "Combo" product, the probability that a user would stop searching jumped from 30.26% to 59.62%. These results were statistically significant (p < 0.01 compared to bulk).
A key takeaway is that the CSA achieved these gains with zero additional resource cost. The agent did not need to find lower prices; it simply changed how it displayed the data it already possessed to better manage user expectations and perceptions of market exhaustion.
Limitations and Ethical Considerations
While effective, the current model assumes an "offline" scenario where the CSA knows all prices before display begins. In real-time systems, the algorithm would need to be adapted to disclose prices as they are gathered. Additionally, the authors note the ethical implications of such psychological "nudging." While it increases CSA revenue, future research is needed to determine the long-term impact on user trust if consumers become aware that their search experience is being strategically manipulated.
Summary updated