Updating Trader Reputation Profiles

Research by Chrysanthos Dellarocas

Reputation mechanisms that allow people to report on their transactions on a public online forum play an important function in the realm of e-commerce, helping to build trust and cooperation among geographically dispersed buyers and sellers. Such mechanisms arose on sites like eBay and Amazon.com because with so many buyers and sellers, the chances were that you would never purchase from the same person twice. Reputation mechanisms inform potential customers of the likely quality of service providers, and they provide an incentive for service providers to do good work. They constitute the digital equivalent of the over-the-fence chatting that used to characterize small-town life.

On a larger scale, business-to-business procurement auctions are becoming more common, and companies such as hospital suppliers are making purchases worth billions of dollars from organizations with whom they have only short-term partnerships. These companies desperately need reputation mechanisms which inform them of the track record of potential business partners, while also giving those partners incentives to provide good service even in the context of a one-time deal.

Information technology has made it possible to control a number of reputation mechanism parameters, including the granularity of solicited feedback, the amount and type of information that is included in a trader’s reputation profile, and the frequency with which reputation profiles are updated with new information. The research of Chrysanthos Dellarocas, assistant professor of decision and information technologies, examines how such design parameters affect trader behavior and market efficiency. His overarching aim is to inform the design of better reputation mechanisms, and thus, of more efficient electronic markets.

Within this broader context, Dellarocas’ recent paper, “How Often Should Reputation Mechanisms Update a Trader’s Profile,” studies the impact of the frequency of reputation profile updates on cooperation and efficiency.

Common wisdom dictates that the more information you have about a seller, the better. Accordingly, most existing reputation mechanisms update user profiles as soon as new ratings are posted to the system. But Dellarocas found that in some circumstances, reducing the frequency at which reputation profiles are updated can increase market efficiency.

Specifically, Dellarocas recommends that a seller’s reputation profile be updated every few transactions with a summary of the seller’s most recent ratings; he finds that there are settings where this induces higher average levels of cooperation, higher seller profits, and higher buyer surplus relative to a mechanism where all ratings are published immediately as they are posted. Why?

Reputation mechanisms rely on self-reporting of transaction outcomes. This opens the door to intentional or unintentional reporting mistakes (reporting noise). Reporting noise reduces the effectiveness of a reputation mechanism because it leads to occasional unfair punishment of honest sellers and, thus, to lower incentives of seller cooperation: if sellers understand that there is a good chance that they might be punished even if they behave honestly they then have a lower incentive to “do the right thing.” In settings where reporting noise is substantial, judging a seller’s behavior on the basis of a batch of reputation ratings, as opposed to publishing each individual rating, minimizes the effects of noise and thus makes it worthwhile for sellers to exert higher effort.

On the other hand, the longer one waits before publishing reputation ratings, the more chances a fraudulent seller has to deceive or give bad service before his behavior becomes public knowledge. Whereas too many updates are subject to noise, too few allow fraudulent behavior to go unpunished. This fundamental tradeoff determines the optimal reputation update frequency. Dellarocas’s paper provides formulae for calculating the optimal update frequency and shows that on several real-life settings the optimal frequency is less than one.

Higher cooperation is good for sellers and buyers, but it also benefits auction operators. In most real-life settings, auction operators charge sellers and/or buyers a percentage of the price paid for each transaction. Given that higher expected cooperation results in higher bids and thus higher auction revenue, a reputation mechanism that maximizes cooperation benefits all stakeholders and, thus, induces more efficient markets.

“More and more people are doing business globally,” says Dellarocas. “When you are evaluating a potential business partner in another country, you desperately need a mechanism to help evaluate that partner. In a small town, if you provided a poor service, the next day the whole town would know about it, but that kind of feedback isn’t possible in a global environment – until Internet-based reputation mechanisms came along. As the value of the items that people buy online increases, so does the need for more effective reputation mechanisms.”

This paper is forthcoming in Information Systems Research and is part of a greater body of analytical reputation mechanism research.