Electronic Commerce Research and Applications
Improving comparison shopping agents’ competence through selective price disclosure☆
Introduction
Comparison shopping is the practice of comparing the prices of items from different sources in order to find the best deal. Yet comparison-shopping is time consuming and requires resourcefulness. In today’s online world, comparison shopping can be substantially facilitated through the use of commercial comparison shopping agents (CSAs) such as PriceGrabber.com, bizrate.com and Shopper.com. These web-based intelligent software applications allow consumers to compare many online stores’ prices, saving their time and money (Pathak 2010). According to Consumer Futures’ report from 2014 (Downer et al. 2013), of the costumers in the UK declared that they have used a CSA in the last 2 years. The 17th annual release of ShoppingBots and Online Shopping Resources (shoppingbots.info) lists more than 350 different CSAs that are currently available online. This rich set of comparison-shopping offerings available over the Internet as well as the fact that each CSA covers only a small portion of the sellers offering a given product, allow prospective buyers to query more than a single CSA for comparison shopping. This way they are more likely to find a good price prior to making a purchase. This poses a great challenge to CSAs, as most of them do not charge consumers for accessing their web sites, and therefore the bulk of their profits is obtained, potentially alongside sponsored links or sponsored ads, via commercial relationships with the sellers they list (most commonly in the form of a fixed payment paid every time a consumer is referred to the seller’s website from the CSA) (Moraga-Gonzalez and Wildenbeest 2012). Therefore, if a CSA could influence buyers to avoid querying additional CSAs, it would certainly improve its expected revenue. In the CSA-buyer setting, the buyer’s decision of whether or not to resume exploration is based primarily on the best price obtained thus far, her expectations regarding the prices that are likely to be obtained through further CSA-querying, and the intrinsic cost of querying additional CSAs (e.g., cost of time). Influencing the best price presented to the buyer can be achieved by increasing the number of sellers whose prices are being retrieved in response to the buyer’s query. Yet, this requires consuming more resources and the expected marginal improvement in the best price decreases as a function of the set size.
In this paper we take a different approach to influence the buyers’ decision whether or not to query additional CSAs, in a way that discourages further querying. The idea is that by disclosing only a subset of all the prices collected by the CSA, one can influence the buyer’s expectations regarding the prices she is likely to encounter if she queries additional CSAs. The underlying assumption is that the buyer is a priori unfamiliar with the market price distribution of the specific product she wants to buy, and her expectations are updated each time she obtains an additional set of prices from a queried CSA (Bikhchandani and Sharma 1996). An intelligent price disclosure strategy can thus decrease the buyer’s confidence in obtaining a better price from the next CSA queried and, as a result, discourage her from any additional querying. We emphasize that this new approach does not conflict with, but rather complements, the idea of increasing the number of sellers whose prices are checked in order to increase the chance of finding a more appealing (lower) price. The paper focuses primarily on situations where the buyer queried a single CSA and needs to decide whether to query more. This is because, as argued later on, this is the common setting, and overall the ability to influence the buyer’s beliefs concerning the market price distribution decreases as a function of the number of prices gathered by the buyer, i.e., the number of CSAs already queried.
The contributions of the paper are threefold. To begin with, we are the first to introduce the idea of selective price disclosure in order to influence buyers to avoid querying additional CSAs. We formally analyze the incentive of buyers to query additional CSAs and CSAs’ benefit in selectively disclosing the prices with which they are acquainted whenever queried. Choosing the best subset of prices to disclose from the original set of prices is computationally intractable, therefore our second contribution is in presenting two price disclosure methods that CSAs can use. These methods are aimed to improve the probability that a buyer will terminate her price-search process and buy the product through the CSA applying the selective price disclosure. Both methods disclose the minimum price known to the CSA, thus the benefit from the partial price disclosure does not conflict with increasing the number of prices the CSA initially obtains to potentially find a more appealing (lower) price. The effectiveness of the methods when the buyer is fully rational is evaluated using real data collected from five comparison shopping agents for four products. The evaluation demonstrates the effectiveness of the resulting subsets of prices achieved with these methods and the tradeoff between their performance and the time they are allowed to execute. Finally, we evaluate the methods using human subjects, to possibly discover that the best solution for fully-rational buyers is less effective with people. This is partially explained by our experimental findings, whereby people’s tendency to terminate their search increases as a function of the number of prices they obtain from the CSA, even if the minimum price remains the same. For the latter population we suggest a simple price disclosing method that has been shown to be highly effective in deterring people from querying additional CSAs.
The rest of the paper is organized as follows: in Section 2 we review related work regarding CSAs, dynamic pricing and selective information disclosure. We formally present the model in Section 3. In Section 4, we analyze the CSA and buyer’s strategies, and the effect of selective price disclosing on the buyer’s decision to query additional CSAs. Later, in Sections 5 Methods, 6 Evaluation with people, 7 When the CSA is not the first to be queried, we discuss and evaluate the selective price disclosing methods for fully rational agents and provide experimental results exemplifying the applicability of the proposed methods with people. Finally, we conclude with a discussions and directions for future research in Section 8.
Section snippets
Related work
The agent-based comparison-shopping domain has attracted the attention of researchers and market designers ever since the introduction of the first CSA (BargainFinder, Krulwich 1996) (Decker et al., 1997, He et al., 2003, Tan et al., 2010). CSAs were expected to reduce the search cost associated with obtaining price information, as they allow the buyer to query more sellers in the same amount of time (and cost) needed to query a seller directly (Pathak, 2010, Bakos, 1997, Wan et al., 2009).
Model
We consider an online shopping environment with numerous buyers (either people or autonomous agents, hereafter denoted “searchers”), sellers and several comparison shopping agents (CSAs), as depicted in Fig. 1.
It is assumed that sellers set their prices exogenously, i.e., they are not affected by the existence of CSAs. This is often the case when sellers operate in parallel markets (Xing et al. 2006), setting one price for all markets. Pricing is assumed to be highly dynamic for the
Individual strategies
In this section, we first analyze the searcher’s optimal search process and the effect of different model parameters on her decision to query additional CSAs. Based on this analysis, we then provide an analysis from the queried CSA’s point of view, discussing the different means available for it to influence the searcher’s search strategy and consequently the CSA’s expected profit.
Methods
Our price disclosure methods are designed in a way that makes them mostly effective in cases where the CSA is the first to be queried by the searcher. This is for two main reasons. First, empirical findings from recent years, indicate that the number of CSAs that searchers query is generally quite modest. For example, a recent consumer intelligence report (Knight 2010) revealed that the average number of CSAs visited by motor insurance buyers in 2009 was . Therefore, when we focus on the
Evaluation with people
While the methods described above are highly effective with fully rational agents, searchers in today’s markets are usually human, and it is well known that people do not always make optimal decisions (Baumeister 2003). In particular, people often follow rules of thumb and tend to simplify the information they encounter. For example, in our online shopping setting, people may ignore the high-range prices rather than use them as part of the distribution modeling (Ellison and Ellison 2009) as
When the CSA is not the first to be queried
The price disclosure methods presented in the previous sections were designed to minimize the critical cost (or termination probability) under the assumption that the CSA is the first to be queried by the searcher. This implied that the prices to be disclosed exclusively influence the searcher’s belief concerning the product’s price distribution. This choice has many motivations, as discussed in detail in Section 5. Still, in some cases the CSA is the second to be queried (or even the kth to be
Discussion and conclusions
The significant increase in searchers’ search termination probability reported in the three preceding sections gives strong evidence for the usefulness of our selective price disclosure approach, both with fully-rational agents and people, in improving a CSA’s expected revenue. As discussed in the introduction, selective price disclosure does not conflict with the general practice of increasing the number of sellers that the CSA queries, as a means for improving the CSA’s competitiveness,
Acknowledgment
This research was partially supported by the Israel Science Foundation (Grant Nos. 1083/13 and 1488/14) and the ISF-NSFC joint research program (Grant No. 2240/15).
References (69)
- et al.
Temporal price dispersion: evidence from an online consumer electronics market
Journal of Interactive Marketing
(2004) - et al.
Optimal search with learning
Journal of Economic Dynamics and Control
(1996) - et al.
Belief extrapolation (or how to reason about observations and unpredicted change)
Artificial Intelligence
(2011) - et al.
Understanding online purchase decision making: the effects of unconscious thought, information quality, and information quantity
Decision Support Systems
(2012) - et al.
Shopbot 2.0: integrating recommendations and promotions with comparison shopping
Decision Support Systems
(2008) - et al.
Modeling the search for the least costly opportunity
European Journal of Operational Research
(2009) - et al.
Truly costly sequential search and oligopolistic pricing
International Journal of Industrial Organization
(2005) - et al.
Dynamic pricing by software agents
Computer Networks
(2000) - et al.
The informational divide
Games and Economic Behavior
(2013) Do lower search costs reduce prices and price dispersion?
Information Economics and Policy
(2005)
Iterated belief change in the situation calculus
Artificial Intelligence
The impact of shopbot use on prices and price dispersion: evidence from online book retailing
International Journal of Industrial Organization
Search and price competition
Journal of Economic Behavior and Organization
A comparison of time-varying online price and price dispersion between multichannel and dotcom dvd retailers
Journal of Interactive Marketing
A personalized and integrative comparison-shopping engine and its applications
Decision Support Systems
Reducing buyer search costs: implications for electronic marketplaces
Management Science
The psychology of irrationality: why people make foolish, self-defeating choices
Persistent price dispersion in online markets
Consumer surplus in the digital economy: estimating the value of increased product variety at online bookseller
Management Science
Negotiated collusion: modeling social language and its relationship effects in intelligent agents
User Modeling and User-Adapted Interaction
Retail strategies on the web: price and non-price competition in the online book industry
Journal of Industrial Economics
When two rights make a wrong: searching too much in ordered environments
Journal of Marketing Resaerch
Search, obfuscation, and price elasticities on the internet
Econometrica
Problem restructuring for better decision making in recurring decision situations
Autonomous Agents and Multi-Agent Systems
Who wins when price information is more ubiquitous? an experiment to assess how infomediaries influence price
Electronic Markets
Reasoning about computational resource allocation
ACM Crossroads
Cited by (17)
On the Reality of Signaling in Auctions
2022, Information (Switzerland)Making money from what you know - how to sell information?
2019, 33rd AAAI Conference on Artificial Intelligence, AAAI 2019, 31st Innovative Applications of Artificial Intelligence Conference, IAAI 2019 and the 9th AAAI Symposium on Educational Advances in Artificial Intelligence, EAAI 2019Adversarial coordination on social networks
2019, Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMASComparison shopping engines: State of the art, exposure of shortcomings and discussion of new innovations
2019, International Journal of Advanced Computer Science and ApplicationsPersonal Loans Comparison Websites in Saudi Arabia: Challenges and Proposed Solution
2018, 21st Saudi Computer Society National Computer Conference, NCC 2018
- ☆
A preliminary version of some of these results appeared in the Proceedings of the Twenty-Seventh National Conference on Artificial Intelligence (AAAI-2013) Hajaj et al. 2013.
- 1
Parts of this work were done when Hazon was at Bar-Ilan University.