Buyers and sellers require trust to make sales on classified sales sites. A failure to establish trust at a sufficiently high level greatly reduces how attractive a listing on a classifieds site is and how fast items are sold. Hence, trust plays an important role in the success of classifieds sales.
In the following article, we cover current practices in measuring credibility and innovation opportunities in tracking and predictive analytics.
In general, one would set up a system that measures credibility to prevent fraud. In particular, eBay and Amazon display the number of transactions and/or the average ratings of profiles. These are extremely useful when evaluating the level of risk because it is unlikely that sellers and buyers with a good record of sales are scammers. However, credibility metrics are built over time, which makes it harder for true buyers or sellers who are starting out and thus have no credibility yet to makes sales quickly and effectively because they are viewed as presenting a higher risk.
One way Airbnb and, to a lesser extent, eBay and Amazon, address this is to allow buyers and sellers to build their own credibility by linking their buyer or seller profiles to other social profiles, e.g., Facebook, Twitter, or Google+. This way, the credibility established on other social networks is transferred to the classified sales site. In addition, Airbnb provides bios, adds phone numbers that have been authenticated with short codes, and verifies emails with activation links.
Combined, these mechanisms greatly help reduce the level of risk of the transaction as perceived by buyers and sellers before the sale and as observed after it is completed.
All these approaches are becoming fairly standard, and they are visible to buyers and sellers. However, behind the scenes, a lot more can and should happen.
Tracking and Predictive Analytics
First, the enrollment process should be properly tracked with analytics so that if a case of fraud is reported on a transaction, a post mortem analysis reveals the fraud markers. Discovering these markers may help the classified site take preventive measures, such as:
- prohibiting some product categories on the website;
- refusing some types of buyers or sellers; or
- closely monitoring transactions known to bear a high risk.
Analytics tends to describe what’s happening, which is already a big step forward, but the game-changing approach is to implement predictive algorithms that are capable of evaluating the level of risk dynamically, before, during, and, in some cases, after transactions. These algorithms can classify or score events based on a series of markers that describe their characteristics. Then, based on the results from these algorithms, i.e., a class or a score, a specific treatment may be defined that says how to “handle” the fraud cases. Typically, three outcomes are possible:
- the transaction is automatically accepted (e.g., white listing);
- the transaction is automatically rejected (e.g., black listing); or
- the transaction is sent for manual verification by a risk operations analyst.
With predictive analytics, there are an infinite number of possibilities for refining how risk is detected, prevented, and handled, which may boost the level of trust on the classified site and, in turn, increase its popularity.
I hope this information helps. As usual, feel free to reach out with questions or comments.
Fabrice | Book