Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q30/00—Commerce
  • G06Q30/06—Buying, selling or leasing transactions

Definitions

• the present invention relates to the sale and/or distribution of goods and services. More particularly, the present invention relates to systems and methods for pricing and/or selling goods or services using randomized sampling and other auction techniques.
• one measure of a digital work's value is the aggregate utility that consumers derive from the work.
• a vendor could recover this amount by charging each individual buyer the maximum amount that each is willing to pay. For example, if a first buyer valued a digitally-recorded movie at $5, a second buyer valued the movie at $8, and a third buyer valued the movie at $1, the vendor could maximize revenue by charging the first buyer $5, the second buyer $8, and the third buyer $1 for a copy of the movie.
• this amount will typically be unattainable in practice, as consumers are generally unwilling to pay a higher price than others for the same item simply because the item is more valuable to them.
• vendors typically estimate consumer utility via market analysis, and then use those estimates to set a fixed price designed to maximize revenue.
• the vendor could maximize revenue by setting the fixed price at $5, thus obtaining a revenue of $10 (i.e., $5 from the first and second buyers).
• Pay-per-view movies are an example of the use of fixed pricing for the sale of digital or electronic content.
• Determining an optimal fixed price can be a difficult task, however, as it requires near-perfect knowledge of consumer utilities. If the price is set too high, an insufficient number of items may be sold; if the price is set too low, insufficient revenue may be collected per item. In the example presented above, if the vendor were to set the price at $6, he or she would only obtain $6 in revenue, as only the second buyer would be willing to purchase the movie. Similarly, if the price were set at $1, the vendor would obtain only $3.
• auctions are sometimes used to determine the sale price.
• An advantage of an auction is that if it is properly designed, it will set the price for an item at or near the optimum fixed price. For example, in a conventional English auction bidders compete against each other to "win" an item at the bid price. Bidders continue raising the bid price until it exceeds the utility value of enough of the other bidders that the number of active bidders is equal to the number of items to be sold. Thus, the winning bidders effectively pay some increment above the utility value of the last bidder to withdraw from the auction.
• Fig. 1 illustrates a Vickrey auction in which ten bidders submit bids for k items. As shown in Fig.
• each bidder has an incentive to bid his or her true utility value, since the price that each winning bid will pay is independent of the value of the winning bid itself.
• unlimited supply refers generally to situations in which the seller has an amount of items that equals or exceeds demand, and/or situations in which the seller can reproduce items on demand at negligible marginal cost.
• unlimited supply refers generally to situations in which the seller has an amount of items that equals or exceeds demand, and/or situations in which the seller can reproduce items on demand at negligible marginal cost.
• the English auction described above were used to sell unlimited supply goods, bidders would have no incentive to raise the price in successive rounds, since all bids would be satisfied no matter what the bid value.
• the Vickrey auction would be ineffective, as each bidder would pay an amount less than or equal to the lowest bid, which the bidders could set at an arbitrarily low level, knowing that it would nevertheless be satisfied.
• One way to avoid these problems is to artificially limit the supply of goods. However, it is apparent that this simply reintroduces the need for market analysis, since the problem of determining how to optimally limit supply so as to maximize revenue is typically no easier to solve than that of
• the present invention provides systems and methods for pricing and selling goods and services of effectively unlimited supply using novel auction techniques. It should be appreciated that the present invention can be implemented in numerous ways, including as a process, an apparatus, a system, a device, a method, or a computer readable medium. Several inventive embodiments of the present invention are described below.
• a system for vending copies of an electronic work includes a computer for receiving bids from one or more bidders, and bid sampling logic for selecting (i) a bid from one of the bidders, and (ii) a sample of one or more of the other bids. Processing logic is used to derive a threshold from the sample of bids, and to determine, based on this threshold, whether to vend a copy of the electronic work to the bidder.
• the system may also include a network interface for sending a copy of the electronic work to the bidder, and payment processing software for collecting payment for the electronic work from the bidder.
• a method for distributing copies of an electronic work. There may be an effectively unlimited number of copies the electronic work, and the maximum number of copies that are to be distributed need not be specified in advance.
• a group of bids is received from one or more bidders.
• a bid is selected from the group, as is a sample of bids drawn from the group of remaining bids.
• a threshold value is derived from the sample, and the selected bid is compared to the threshold. If the selected bid is greater than or equal to the threshold, a copy of the electronic work is sent to the bidder who submitted the selected bid. This process can be repeated for each bid in the group.
• a method for distributing an electronic work is disclosed.
• a group of bids are received, and a weighting factor is assigned to each bid.
• a first bid is selected from the group of bids, and another bid is drawn from the group using the weighting factors.
• the weighting factors can be used to make the probability of drawing a given bid from the group proportional to the relative value of the bid in comparison to the other bids in the group.
• the value of the first bid is compared to the value of the bid that was drawn from the group, and a copy of the electronic work is distributed to the bidder who submitted the first bid if the first bid is greater than or equal to the other bid.
• Fig. 1 illustrates the operation of a Vickrey auction.
• FIGs. 2A, 2B, and 2C illustrate a system for practicing an embodiment of the present invention.
• Figs. 3A, 3B, and 3C are flow charts illustrating sampling-based optimal threshold auction mechanisms in accordance with embodiments of the present invention.
• Figs. 4A and 4B illustrate the application of a dual-price, optimal threshold auction mechanism to a set of bids.
• Fig. 5 is a flow chart illustrating a method for conducting an auction in accordance with an embodiment of the present invention.
• Figs. 6A and 6B illustrate the application of a weighted-pairing auction mechanism to a set of bids.
• Fig. 7 illustrates the performance of exemplary auction mechanisms when bids are distributed uniformly.
• Fig. 8 illustrates the performance of exemplary auction mechanisms when bids are distributed according to a normal distribution.
• Fig. 9 illustrates the performance of exemplary auction mechanisms when bids are distributed according to a Zipf distribution.
• Fig. 10 illustrates the performance of exemplary auction mechanisms when bids are distributed according to an equal-revenue distribution.
• Fig. 1 1 illustrates the performance of exemplary auction mechanisms when bids are distributed in a bipolar fashion.
• a randomized auction mechanism is used to determine both the number of goods that are sold and the selling price.
• the auction mechanism automatically adapts to the bid distribution to yield revenue that is competitive with that which could be obtained if the vendor were able to determine the optimal fixed price for the goods.
• a set of bids is randomly or quasi-randomly partitioned into two or more groups. An optimal threshold is determined for each group, and this threshold is then used to select winning bids from one or more of the other groups.
• each bid is compared to a competing bid that is randomly or quasi-randomly selected from the set of bids. If the bid is less than the randomly- selected competing bid, the bid is rejected. Otherwise, the bid is accepted and the bidder buys the auctioned item at the price of the randomly-selected bid.
• system 200 preferably includes:
• bidders submit bids electronically via network 214 to server 211.
• Server 211 is operable to send the bidders a list of the items being auctioned and to facilitate entry of bids from terminals 210.
• Server 211 is also preferably operable to identify the winning bids, to accept and/or process payment for the auctioned item, and to make and/or dispense a copy of the auctioned item to each of the winning bidders.
• server 211 might comprise a mainframe computer or minicomputer running Microsoft® Site Server 3.0, Commerce Edition software, available from Microsoft Corporation of
• server 211 may include:
• a network interface 230 for connecting server 211 to network 214 and for communicating with terminals 210; • a magnetic and/or optical disk drive 232 for reading and writing diskettes and/or CDs containing content and program files;
• a non-volatile storage unit 234 such as a hard disk drive, for storing content and program files
• RAM random access memory
• ROM read only memory
• CPU central processing unit
• server 211 The operation of server 211 is controlled primarily by control programs that are executed by the server's CPU 244. These control programs may be stored in system memory 236. In a typical implementation, the programs stored in system memory may include:
• a security module 253 for securely transmitting and receiving data to and from network 214, display 240, and/or disk drives 232 and 234; • an auction control module 254, discussed in more detail below, for managing an online auction, preferably including modules for transmitting information to bidders; receiving bids; evaluating bids to identify winners and losers; making copies of digital content and transmitting those copies to the winning bidders; obtaining and processing payments from the bidders; recording and analyzing bid distributions; allowing the auctioneer to select the auction mechanism to be used; and responding to auctioneer and/or bidder queries regarding the nature of the auction.
• terminal 210 may comprise any suitable mechanism for communicating a bid to server 211.
• terminal 210 may comprise a personal computer 210a, a personal digital assistant (PDA) 210b, a television with set-top box 210c, or the like.
• terminal 210 might include: • a user interface 274, including a display 276 and one or more input devices 278;
• a network interface 280 for connecting terminal 210 to network 214 and for communicating with server 211;
• storage media 284 operable to store digital content downloaded from server 211;
• system memory 286 for storing digital content and control and security programs that govern the operation of the system and its interaction with network 214 and input/output devices 278;
• communication between server 211 and terminals 210 is performed using a secure protocol such as the Secure Sockets Layer (SSL) protocol, the Internet Engineering Task Force's (IETF) Transport Layer Security (TLS) protocol, and/or the secure HTTP (S-HTTP) protocol to create an effectively secure channel between the server and each terminal and/or to send individual messages securely.
• SSL Secure Sockets Layer
• IETF Internet Engineering Task Force's
• TLS Transport Layer Security
• S-HTTP secure HTTP
• SSL can be used by server 211 to verify the identity of a terminal 210 (and/or vice-versa) before valuable content or sensitive financial information is transmitted, and can also be used to facilitate encryption of the information that is ultimately transmitted.
• information is sent between terminals 210 and server 211 using the methods and systems described in the '900 patent, previously incorporated by reference herein.
• a user's bid can be enclosed in a secure container and sent to server 211, where the secure container is opened and the bid removed.
• server 211 may be operable to send the auctioned item to each winning bidder in a secure container, such as a DIGIBOX ® secure container produced by InterTrust Technologies Corporation of Santa Clara, California.
• a preferred embodiment uses the systems and methods described, e.g., in the '900 patent, the Sibert application, and or Menezes et al., Handbook of Applied Cryptography, pp.
• Fig. 2A illustrates a system 200 in which bids are entered remotely at computer terminals 210 and forwarded via network 214 to server 211 for processing and storage
• bids can be submitted via a variety of additional and/or alternate means, including without limitation facsimile, telephone, email, the postal system, and/or hand delivery. These bids could then be entered manually into server 211 by the auctioneer and/or processed manually.
• the system for processing bids shown in Figs. 2A, 2B, and 2C is intended to be illustrative and not limiting.
• auction control module 254 includes a variety of bid processing procedures 258 for conducting an auction (or lottery) in accordance with the principles of the present invention. Several illustrative procedures will now be described with reference to Figs. 3A - 6.
• an optimal threshold for a randomly-selected sample of the bids obtained by server 211 is used to determine the outcome of some or all of the remaining bids.
• the optimal threshold of the sample is one that would maximize the revenue obtained from the sample.
• Fig. 3A is a flow chart illustrating one such sampling-based threshold technique. With reference to Fig. 3A, after gathering a set of bids (310), a sample group of the bids is selected (312). In a preferred embodiment, a pseudo-random sampling of the bids is taken, with the size of the sample being chosen such that the properties of the sample can be expected to be statistically representative of the properties of the entire group of bids. For example, for a relatively large number of bids, n, sample sizes of approximately the square root of n or H/10 have been found to work well. However, it will be appreciated that other suitable sample sizes could be used in accordance with the principles of the present invention.
• an optimal threshold for the sample is determined (314), the threshold being a value that, e.g., maximizes the revenue that could be obtained from the sample.
• the bids in the sample are then discarded (315). That is, none of the bidders in the sample are allowed to purchase the item, regardless of the price that they bid. This has the beneficial effect of making the price that the winning bidders pay independent of the value that they bid, thus encouraging each bidder to bid his or her true utility value.
• the threshold from the sample is then compared against the values of the bids outside the sample (316 - 326). If a bid value is greater than (or in some embodiments, greater than or equal to) the threshold (a "yes" exit from block 320), then server 211 sends a copy of the item to the bidder (322) and collects payment from the bidder in an amount equal to the threshold (324). If, on the other hand, the bid is less than or equal to (or in some embodiments, less than) the threshold (i.e., a "no" exit from block 320), the bid is rejected (318). In a preferred embodiment, when a bid is rejected a message is sent to the bidder indicating that the bid was unsuccessful. This process is repeated for each of the bids in the remainder.
• a dual-priced technique is used to avoid the loss of revenue that results from discarding the bids used to set the threshold.
• Figs. 3B and 3C illustrate this technique.
• the bids in the sample are not discarded. Instead, a second threshold is determined by analyzing the bids in the remainder (348). This threshold is then used to determine whether the bids in the original sample are winners or losers (350 - 360).
• the threshold used to select each winning bid is chosen independently of the value of the winning bid itself, and thus bidders are incentivized to bid their true utility values.
• the sample is taken such that it is approximately equal in size to the group of unsampled bids (i.e., the sample and the remainder are of approximately equal size).
• the order of the blocks shown in Figs. 3 A, 3B, and 3C can be varied without departing from the principles of the present invention.
• blocks 318, 322, and 324 could simply comprise the act of recording the status of b,- (e.g., win or lose), with all of the bids being processed together at the end of the auction (e.g., rejection notifications sent, copies of the auctioned item made, and/or payments processed). It will also be appreciated that in some embodiments, some steps need not be performed.
• payment may not need to be collected.
• payment can be collected (or at least reserved) when the bids are initially received - payment of a deposit, or demonstration of the ability to pay, being a prerequisite to having a bid considered. In such embodiments, funds can simply be released to the vendor once a bid wins.
• Figs. 3B and 3C can be generalized to the case where an arbitrary number of samples are taken from the original set of bids, thresholds for each of the samples are computed, and thresholds from one sample (or set of samples) used to evaluate which bids in another sample win or lose. Accordingly, it will be appreciated that there are numerous ways to apply the techniques disclosed herein, any suitable one or more of which could be used in accordance with the principles of the present invention.
• the original set of bids could be divided into two groups in numerous different ways: at one extreme, a single bid could be selected at random, and its value used as a threshold against which the other bids are compared, while at another extreme, each bid could be compared to the optimal threshold of a "sample" consisting of the entire set of remaining bids.
• Figs. 4A and 4B illustrate the application of a dual-priced, random-sample, optimal threshold technique to a group of twenty bids.
• the bids are randomly (or arbitrarily) partitioned into two groups, shown as a shaded group and an unshaded group.
• the optimal threshold for each group is calculated.
• the optimal threshold for the shaded group of bids i.e., the threshold that would yield the greatest revenue if applied to the shaded group
• This threshold is applied to the bids in the unshaded sample, yielding a revenue of $18 (i.e., all unshaded bids of $2 or greater pay $2).
• the optimal threshold for the unshaded sample is determined - in this case it is also $2 - and applied to the bids in the shaded sample, yielding a revenue of $16.
• the auction mechanism is able to achieve the same revenue as an optimal fixed pricing scheme, yet by using the auction mechanism the vendor is able to avoid the costs of performing the market research necessary to determine the optimal fixed price. It can be shown that if, for example, the highest bid is small in relation to the total amount of revenue obtainable through optimal fixed pricing, then the expected revenue of a random sample optimal threshold auction will be within a constant factor of the revenue of the optimal single-price auction, and will usually be very close to the revenue obtained using fixed pricing with perfect market information.
• random is intended to encompass pseudo-random, quasi-random, effectively random, or arbitrary processes, in addition to “truly” random processes.
• Any suitable random number generation technique can be used in accordance with the teachings of the present invention, including without limitation those set forth, in Knuth, The Art of Computer Programming: Volume 2, Seminumerical Algorithms, 3d ed., pp. 1-193 (Addison- Wesley 1998), which is hereby incorporated by reference.
• Fig. 5 is a flow chart illustrating another illustrative method of conducting an auction for digital goods.
• an auction mechanism is implemented where for each bid x, a different bid is selected using a predefined selection process that is effectively independent of x. liy is less than or equal to x, the bidder who bid x wins and pays y. Otherwise the bidder loses
• a set of bids is gathered (510) and weighted using a suitable weighting scheme (512).
• the bids are weighted proportionally to their bid value; however, one of ordinary skill in the art will appreciate that other weighting schemes can be used. For example, the weighting scheme need not be linearly proportional to bid value. In other embodiments, no weighting scheme is used.
• each bid, b is compared to a randomly selected competing bid, bk (513 - 524).
• Figs. 6A and 6B illustrate the application of the weighted-pairing auction mechanism desc ⁇ bed above to a set of ten bids.
• m a preferred embodiment each bid is assigned a weight proportional to the value of the bid. In the embodiment shown in Fig. 6A, this is accomplished by assigning each bid, b congestion a number equal to the sum of all the bid values through b,.
• bids can be randomly selected with a probability proportional to the individual bid values simply by randomly generating a number between 1 and ⁇ bid_values (in Fig. 6 A, a number between 1 and 30) and mapping that number back to its corresponding bid. For example, with the weighting scheme shown m Fig.
• a weighted pool of bids could be created by duplicating each bid a number of times proportional to the value of the bid, then selecting each bk at random from the resulting pool.
• the weighting is performed separately for each b, - that is, for each _>, the remaining bids are weighted - which is equivalent to relocating weighting block 512 in Fig. 5 such that it is located between blocks 513 and 514.
• Fig. 6B illustrates the result of applying the process shown m Fig. 5 to the bids shown in Fig. 6A.
• b k is randomly selected using the weighted selection scheme desc ⁇ bed above.
• b k is selected from the set of all bids except for b congestion which can be accomplished by simply rejecting random numbers that map to b, rather than to a unique b k -
• Each _>, is compared to its corresponding b k , and a determination is made as to whether b, wins or loses based on whether b, is greater than or equal to b k (or, in other embodiments, whether b, is greater than bk).
• the vendor determines the identity of the winning bidders in this manner, collects payment equal to the approp ⁇ ate value of b k for each bid, and sends a copy of the auctioned item to each winning bidder.
• the weighted pairing auction desc ⁇ bed above is multiple-p ⁇ ced and stable. Moreover, as shown Figs. 7, 8, 9, 10, and 11, theoretical and experimental results indicate that this auction performs relatively well in comparison to other stable auction mechanisms.
• One way to improve the performance of the weighted pairing mechanism for some bid distributions is to pick a constant, g, and to modify the auction so that _>, ⁇ wins if it is greater than g*bk, and pays g*b k -
• g is a constant between 0 and 1.
• the modified auction revenue will be at least g times the original auction revenue, but has the potential to be much larger.
• Figs. 7, 8, 9, 10, and 11 compare the performance of optimal fixed pricing (i.e., the performance that could be obtained with perfect market analysis) to the performance of several embodiments of the present invention. In these plots, the performance of various bid mechanisms - measured on the y-axis as a percentage of the revenue obtainable by optimal fixed pricing (F) - is plotted versus the total number of bids in the auction. Curves for the following auction mechanisms are illustrated in Figs. 7, 8, 9, 10, and 11:
• DOT Deterministic Optimal Threshold
• Fig. 7 illustrates the performance of the exemplary auction mechanisms when bids are distributed uniformly (e.g., randomly) between a maximum bid of 1 and a minimum bid of 0.
• Fig. 8 illustrates performance when bids are distributed according to a normal distribution curve with mean of 1 and standard deviation of 1 (negative bids are not included).
• the random sampling auction mechanisms perform particularly well on the uniform, normal, and Zipf distributions, as these distributions are characterized by the fact that a uniformly-chosen random subset of the bids will, on average, have the same distribution as the original distribution.
• the auction mechanisms of the present invention perform better than imperfect fixed pricing.
• Fig. 10 the distribution is such that the same revenue can be obtained by setting the selling price to any bid value (except for the highest) and satisfying all bids greater than or equal to that value. As seen in Fig. 10, this bid distribution represents a worst-case performance for several of the auction mechanisms.
• Fig. 11 illustrates the performance of various auction mechanisms when bids are clustered in a bipolar fashion (i.e., bids are clustered at high and low values). In Fig. 11, the performance of the auction mechanisms are plotted for various bid-cluster configurations.
• the information obtained by the auctioneer in a particular auction can be useful in determining the optimal method by which to sell additional copies of the goods in the future. For example, if the auction mechanism encourages bidders to bid their true utility values (i.e., is a stable auction mechanism), the auctioneer will collect valuable data on consumer utility values and the distributions thereof. Using these data, the auctioneer can adapt future sales mechanisms to best match the distributions of utility values observed in the market for the goods in question.
• the auctioneer can conduct future auctions for this type of goods using a variant of the weighted-pairing auction mechanism or the single-sample, optimal threshold mechanism, both of which have been found to perform well on such a distribution.
• the vendor could simply set a fixed price at the optimal level derived from the auction data.
• the vendor could capitalize on consumer price discrimination by offering a modified form of the goods which would be less appealing to consumers who place a high utility value on the item, but which would still be appealing to consumers who place a low utility value on the item.
• the vendor could create a version of the goods that did not include some of the features or functions valued by the high-end customers.
• the vendor could then hold separate auctions for each class of goods, setting a reserve price on the high-end version of the goods, the reserve price being greater than the utility values exhibited by the consumers of the low-end version.
• a stable auction mechanism enables the vendor to accurately observe market distributions and to adapt in an optimal fashion.
• auction application software 254 includes data analysis modules 262, which are operable to record the bid distributions observed in actual auctions and to determine an optimal auction mechanism and or fixed price for the observed bid distribution. In one embodiment, this determination is performed by stepping through a library of auction mechanisms, and iteratively optimizing the auction variables to maximize the revenue derived from the auction. For example, the sample size of the single-price optimal threshold mechanism could be optimized; the choice of scaling factor g could be optimized; and so forth.
• the dual-price optimal threshold mechanism can be adapted to an auction for k items by simply selecting the optimal threshold for each sample to yield approximately k/2 winning bids. If too many bids are selected, bids can be randomly rejected until only k bids remain.
• auction mechanisms presented herein can be similarly adapted for use in the context of limited supply goods.
• the systems and methods described herein are readily adaptable to situations where this is not the case. For example, if the marginal cost of producing an item is v, the vendor can simply subtract v from all bids, automatically reject all negative bids, and then conduct the auction using the remaining bids in the manner previously described (with v added back to each bid). This technique can thus be used to enforce a vendor's "reserve" price.

Landscapes

• Business, Economics & Management (AREA)
• Accounting & Taxation (AREA)
• Finance (AREA)
• Development Economics (AREA)
• Economics (AREA)
• Marketing (AREA)
• Strategic Management (AREA)
• Physics & Mathematics (AREA)
• General Business, Economics & Management (AREA)
• General Physics & Mathematics (AREA)
• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=22555514

Family Applications (1)

Country Status (3)

Families Citing this family (1)

Citations (3)

• 2000
  • 2000-09-21 AU AU40209/01A patent/AU4020901A/en not_active Abandoned
  • 2000-09-21 WO PCT/US2000/025763 patent/WO2001022320A2/en not_active Ceased
  • 2000-09-21 EP EP00963655A patent/EP1222589A2/de not_active Ceased

Patent Citations (3)

Non-Patent Citations (4)

Also Published As

Similar Documents

Legal Events