Ben's Website

Serious musings

PreReview

marketsopinionAmazoneBay

I was reading this article which mentions review-buying and ended up re-reading my post on "If I Were eBay" and had some additional thoughts.

My conclusion there was:

These are all free features of online commerce and, rather than overburdening the customer, are actually ways to decrease customer purchase risk while also learning more about the market dynamics. This is a very crucial point: the customer's frustration in composing search queries is mirrored by the seller's frustration in understanding why those customers eventually bought what they did. Markets are complex, it is eBay's job to help both seller and buyer understand each other and come to terms. To the degree that customers are unable to evaluate and understand product risk on Amazon, there is room to compete. So, if I were eBay I would focus on users, explaining market dynamics and telling them how we've engineered our marketplace to maximize their satisfaction.

Save me from wading through pages of crappy knockoffs to find the thing I'm actually looking for, and I might not mind the absence of 2-day shipping! Voila. The job of the marketplace is to decrease purchase risk, not to increase it by obscuring actual selection and hiding real differences.

Stepping back, the way that most online markeplaces work is that the customer discovers their need for something, does some research on/off-site to better understand their need and what might satisfy it, eventually buys an item, receives it, uses it, and is occassionally prompted to leave a review, whereupon this journey might be narrated. There are many motivations for customers to leave reviews, but they are unequal, leading to reviews by only a subset of customers with highly variable motivation and irregular quality. All useful reviews can be seen as contributing information lacking in the product description provided by the seller.

Can the situation of product listings and reviews be improved in 2025? One observation is that the a generic review is composed of:

  1. initial expectations about the product performance, that the purchase is intended to solve a particular set of problems,
  2. results thereof, and
  3. experiences with the product that were unknown or unpredicatable before use.

Reviews mostly solicit 2), because they are often written in-the-moment of ire or joy. But 1) is worth soliciting as it encompasses market evaluation, how various customers compare various products.

Now that we can have AI shopping assistants, 1) becomes a conversation that is beneficial to the degree that it accurately summarizes the products and their comparative value. Pulling in external reviews is an initial source of content, but the true value emerges post-purchase, where the platform can incent review conversations. In these the AI would ask several gap-filling questions of the customer, with follow-ups, with the goal of informing the pre-purchase conversation of this and other customers. This would more natually and even-handedly generate 2) and 3), while greatly if temporarily preventing review fraud.