First, we explore what happens when MY's re-pricing rate is increased. Intuitively, it is clear that such a pricebot would spend more of its time undercutting its competitors. Analysis and simulations confirm that, if all sellers adopt MY, then the one that re-prices fastest makes the most profit. (This observation seems likely to extend to other pricing strategies as well.) An arms race in re-pricing rate seems inevitable.
Let us develop this scenario a little further. Suppose that a bookseller like amazon were to implement a pricebot that reset prices on one million titles every day. In order to compute the new prices using MY or a related strategy, amazon would have to obtain quotes from each competitor on each title. To do this, amazon could use its own special-purpose shopbot. However, it might prefer to write a simple agent that automatically sends requests for price quotes to DealPilot or another commercial shopbot. This would take much less effort, and it would also hide the fact that amazon was seeking valuable price information from its competitors. If amazon requests one million price comparisons per day, each of which takes 20 seconds on average, the shopbot would be hit with several hundred requests simultaneously.
For this herculean task, the shopbot would gain absolutely nothing. Many shopbots earn their living by selling advertising space on their Web pages or by making a commission on sales made through referrals. However, amazon would not make any actual purchases from its competitors, so DealPilot would get no commissions, and amazon's agent would not read any of the ads put up by DealPilot.
The scenario gets worse. Assume that DealPilot is somehow able to meet amazon's constant demand. In order to gain a competitive advantage over amazon, bn.com might ask DealPilot to check prices on one million titles every hour. Then borders might up the ante to one million titles per ten minutes, and another dozen online booksellers could follow suit. DealPilot could wind up receiving over a billion requests per day! It is doubtful that such a load could be handled. Even if the booksellers were relatively conservative and just focussed on the ten thousand most popular titles, the load on DealPilot could easily dwarf the number of requests coming from legitimate buyers. Keep in mind that each request for a price comparison from a bookseller translates into several requests for price quotes that the shopbot submits to booksellers. Thus the total network traffic and the hit rate on each bookseller's server is potentially enormous.
A reasonable solution to the problem of excess demand for shopbot services would be for shopbots to charge pricebots for price information. Even if the cost per price-comparison were just a fraction of one cent, one might expect to reach a balance point beyond which the benefit of requesting extra price comparisons would be less than the cost of obtaining these comparisons from the shopbot. Once shopbots begin charging for pricing information, it would seem natural for sellers -- the actual owners of the desired information -- to themselves charge shopbots (and possibly other clients) for their information. The sellers could use another form of pricebot to dynamically price their own price information. This illustrates how dynamic pricing of information services could quickly percolate through an entire economy of software agents. We expect to see such situations repeatedly as the Internet makes the transition to the Information Economy: services or agents that charge for their information goods or services will create the proper incentives to encourage business but deter excessive, counterproductive resource usage.
Since it is pointless and inefficient to change the price of an item more frequently than it is being requested by buyers, another likely development is that sellers will price their wares only on demand. This is precisely what Books.com does today -- it dynamically sets a price on a book when a buyer expresses interest. However, suppose for example that borders were to adopt the same policy as Books.com. When Books.com checks borders' price on a book, borders does not have a ready-made price; instead, it must check other booksellers' prices -- including Books.com's! Unless some care is taken, an endless pricing loop could be generated. One way to avoid such problems would be to set up a reverse auction whenever a buyer expresses interest in a book. Booksellers would bid against one another, and the lowest bidder would be obligated to sell the book to the buyer for the agreed-upon price. Thus the natural limit of ever-faster nonnegotiable dynamic pricing would appear to be negotiable dynamic pricing mechanisms such as auctions, in which the dynamics may occur over the course of the transaction.