Research and Advances
Artificial Intelligence and Machine Learning Virtual extension

Technical Opinion: Online Auctions Hidden Metrics

  1. Introduction
  2. Listing Errors
  3. Seller Frustration
  4. Auction Fraud
  5. References
  6. Authors
  7. Footnotes
  8. Figures
  9. Tables

Most research articles to date in online auctions focus on the study of auction performance, such as auction success and ending bid price. Popular factors under scrutiny include starting bid, bid increment, auction duration, number and quality of the displayed pictures, seller and buyer feedback rating, payment method, buyer protection, among others. Clearly, these are what can be directly manipulated by sellers or observed by researchers. However, the number of online auctions that don’t result in valid transactions is astonishing, which leads us to call for a better understanding about unsuccessful auctions. Why are there so many of them and what can be done about it?

The online auctions literature has put disproportionate emphasis on explaining the bidding process, rather than the selling process. The focus of most research articles has been on explaining bid formation and bidder behavior. In practice, however, selling goods in a marketplace like eBay can be complicated for a novice seller, and cause a great deal of frustration, as so many parameters need to be appropriately chosen and all the necessary information listed.

We believe that there are important hidden metrics that directly affect sellers and are mostly overlooked by researchers. A better understanding of them can lead to an improved auction mechanism design that ends in a win-win-win situation for the auction house, sellers, as well as buyers. These metrics may not seem critical in the first glance but in actuality contribute to a significant extent the auction performance. The hidden metrics we are reporting here include auction listing errors, seller frustration, and auction fraud. As we present below, these metrics have a strong impact on all sellers, in particular inexperienced sellers.

We compare between relatively new sellers and more experienced sellers in each of the metrics. The intent is to uncover the correlations between seller experience and these metrics. Since the auction literature provides supporting evidence for better performance by experienced sellers in terms of their positive feedback ratings, findings in this study shed light in how differently sellers perform in these hidden metrics that might have contributed to an asymmetrical overall auction performance. It will be able to provide some guidance for new sellers in improving the outcome of their auctions.

Data used in this study include 23,355 auctions in eBay for Sony PlayStation 2 console. Data collection period spanned three and half months from June 2, 2005 to September 17, 2005. That is, all Sony PlayStation 2 eBay auctions that ended in the 3½-month period were included in our study. We chose Sony PlayStation 2 because it dominated the video game console market at the time of the data collection, indicating few interfering noises in our analysis. In addition, due to its market dominance, it has reasonable market thickness measured by number of listed auctions and daily bids.

Back to Top

Listing Errors

Listing errors refer to seller mistakes in setting parameters for an auction, for example, mistakenly setting an extremely low reserve price. According to eBay policy, if a seller finds any listing error during the auction period, she has the right to cancel the auction. This cancellation action might frustrate bidders if bids have already been placed and the listing error(s) is to the bidders’ advantage. On the other hand, if the seller decides to continue the auction with the error(s) and honors the wrong listing, she might incur significant costs.

Within the 3 ½-month data set, we observed 362 listing errors that led to auction cancellations. Interestingly, we found that the error rate is roughly inverse proportional to the seller’s feedback scores (ratings). That is, newer sellers are more likely to incur listing errors in their auctions. For example, sellers with 0 feedback scores occupy 4% of the total seller population but this group’s listing errors accounts for 8.3% of the total errors. 50% of all listing errors are caused by sellers with feedback scores of less than 34, but these sellers only make up 35% of the total seller population. We further conducted a statistical test and the result indicates that it is statistically significant that relatively new sellers incur higher auction listing errors.

Listing errors can potentially lead to major frustration of sellers and/or buyers. In the worst scenario, significant loss might occur. From our data observation, it seems that newer sellers are the main source for these listing errors. To improve the situation and reduce auction cancellations and frustrations, better tutorial systems need to be in place for new sellers to easily learn the overall process of strategically setting up an auction. In the long run, the costs of implementing such systems will be offset by the better percentage of successful auctions and happier market participants.

Back to Top

Seller Frustration

Not all auctions end successfully. In the data set for this study, the average unsuccessful rate is 19% for the 23,355 auction listings. Possible causes for unsuccessful auctions include no bids at all, reserve price not met, or seller cancellation. For whatever reason, if an auction did not end successfully, the seller can elect to withdraw the selling item from the market completely or re-list the item for another auction. We measure sellers’ frustration through the extent of their withdrawing selling items from the market after failed auctions. According to our definition, the lower this withdrawing rate the less frustrated the sellers. An implicit assumption here is that if a seller decides to re-list an item, he/she will do it within a month of the failed auction. We believe this is a reasonable assumption. Even in the rare cases where some sellers wait for more than a month to re-list an item, our analysis should still be valid since the situation are likely to occur evenly among since these situations are new and experienced sellers. Under this assumption, we used the first month and last month as controlling windows, and observed the 1,242 unsuccessful auctions after the first-time listing in the middle one-and-half months (between July 2, 2005 and August 17, 2007) as a base to figure out the withdrawing rates.

Table 1 shows the percentage of items withdrawn from the market after the first-time auction failure for both new and experienced sellers. Following prior research, we use seller feedback scores as a proxy for seller experience. Since the definition of new seller is fairly subjective, several values of threshold feedback scores were used and results displayed. Clearly, new sellers tend to withdraw items from the market more than experienced sellers after failed auctions. A simple Z-test shows that the result is statistically significant at 0.01 level for all of the new seller definitions used. To demonstrate the distribution of seller frustration, Figure 1 shows the histogram of seller frustration for different feedback scores.

If an item is withdrawn from the market, the auction house loses the potential commission fee from selling it and the seller loses the revenue that can be accrued if the item is re-listed and sold. It is obvious to the advantages of the auction hosting house and the seller that a failed item is re-listed. More efforts should be put into helping sellers choose appropriate auction selling parameters, and revise their previous listings, especially the ones without much experience so that their confidence increases and the outcome of the auction improves.

Back to Top

Auction Fraud

Online auction fraud poses great concerns to market participants simply because most online transactions are based on trust. Significant efforts have to be in place to suppress any kind of auction fraud so both sellers and buyers feel safe to transact online. In Figure 2 we report the distribution of auction fraud claims for feedback scores. We manually examined the 381 total negative feedback received among all sellers during the study period and found that buyers claimed a total of 66 fraud claims. The histogram shows that approximately 50% of all fraud claims are caused by sellers having a rating higher than 15. A further investigation reveals that these sellers comprise less than 22% of the total seller population. This indicates that newer sellers tend to generate more fraud claims by buyers than the experienced sellers. Figure 2 also shows that in addition to low experience levels, sellers who are recent members of eBay are responsible for the highest share of fraud claims. An immediate implication is that honest new sellers might undergo a tremendous pressure of winning buyers trust due to some dishonest new entrants.

We also collected information from eBay about seller account suspensions. Out of 12,289 unique sellers, we observed 467 suspensions. A detailed analysis of these account suspensions reveals a close similarity distribution between account suspension and auction fraud claim. That is, sellers with less feedback scores have more auction fraud claims and more account suspensions. We did not analyze buyer auction fraud since most successful auctions require that the sellers first receive payment before the item is shipped. Therefore, buyer auction fraud is, in our opinion, lesser in severity with regard to damages.

In this article, we report three auction metrics, namely, auction listing errors, seller frustration, and auction fraud, which do not typically appear on the radar screen of most auction research. In our opinion, however, these metrics deserve more attention as their impact might be far greater than what’s generally perceived or even ignored. We present some preliminary results of data obtained from real online auction setting. Based on the findings about listing errors and frustration rates, it is clear that the auction houses need to pay more attention to helping novice sellers. Tutorials and online recommendation systems should be in place to minimize listing errors and to help novice sellers make the right selection of auction parameters that can enhance the probability of achieving successful transactions. Follow-up to sellers who were not successful in the early offerings should provide support and guidance for their continuing engagement.

As for the fraud tendencies we uncovered, they indirectly hurt inexperienced sellers, as it seems that fraud-inclined sellers keep changing their registration and resurface as new sellers. The auction house needs to be more proactive in raising the awareness about these occurrences to both buyers and sellers. Overall, a more careful investigation of these metrics, as discussed earlier, can lead to a win-win-win situation for the auction house, sellers, as well as buyers.

Back to Top

Back to Top

Back to Top

Back to Top


F1 Figure 1. Seller Frustration Measure Distribution.

F2 Figure 2. Auction Fraud Claims Distribution.

Back to Top


T1 Table 1. Measurement of Seller Frustration.

Back to top

    1. Anderson, S., Friedman, D., Milam, G., and Singh, N. Seller strategies on eBay. Economics Working Paper Department of Economics, University of California, Santa Cruz, 2004.

    2. Balachander, S. Warranty signaling and reputation. Management Science 47, 9, 2001, 1282–1289.

    3. Ba, S. and Pavlou, P. Evidence of the effect of trust building technology in electronic markets: Price premiums and buyer behavior. MIS Quarterly 26, 3, 2002, 243–268.

    4. Bapna, R., Goes, P., and Gupta, A. Replicating online yankee auctions to analyze auctioneers' and bidders' strategies. Information Systems Research 14, 3, 2003, 244–268.

    5. Bapna, R., Goes, P., Gupta, A., and Jin, Y. User heterogeneity and its impact on electronic auction market design: An exploratory analysis, MIS Quarterly 28, 1, 2004, 21–43.

    6. Becherer, R and Halstead, D. Characteristics and internet marketing strategies of online auction sellers. International Journal of Internet Marketing and Advertising 1, 1, 2004, 24–36.

    7. Dewally, M. and Ederington, L. A Comparison of reputation, certification, warranties, and disclosure as remedies for information asymmetries: Lesson from the on-line comic book market. Journal of Business 79, 4, 2006.

    8. Gilkeson, J. H. and Reynolds, K. Determinants of internet auction success and closing price: An exploratory study. Psychology & Marketing 20, 6, 2003, 537–566.

    9. Hu, X., Lin, Z, Whinston, A. B., and Zhang, H. Hope or hype: On the viability of escrow services as trusted third parties in online auction environments. Information Systems Research 15, 3, 2004, 236–249.

    10. Sinha, A.R. and Greenleaf, E. A. The impact of discrete bidding and bidder aggressiveness on sellers' strategies in open english auctions: Reserves and covert shilling." Marketing Science 19, 3, 2000, 244–265.

    Work supported by CIDRIS – Center for Internet Data Research and Intelligence Services at OPIM / University of Connecticut.

    Part of this article was written while Dr. Tung visited National Cheng Kung University (NCKU) in Taiwan as a Joint Assignment Associate Professor in September, 2006.


Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More