| CPC G06Q 30/0275 (2013.01) [G06Q 30/0277 (2013.01)] | 47 Claims |
|
1. A system for a hybrid, optimized exchange operably connected to a demand-side platform (DSP), the hybrid, optimized exchange further operably connected to a supply-side platform (SSP), the hybrid, optimized exchange configured to receive an advertising request, the hybrid, optimized exchange further configured to conduct, using a bid floor, a hybrid, optimized exchange DSP auction of the advertising request among a plurality of DSPs, thereby generating a winning DSP that makes a winning DSP bid in an automated advertising auction system after the SSP initiates an SSP auction of the advertising request, the hybrid, optimized exchange optimizing the bid floor provided to the plurality of DSPs in the hybrid, optimized exchange DSP auction and simultaneously optimizing a shading factor used by the hybrid, optimized exchange to place a hybrid, optimized exchange bid on behalf of the winning DSP in the SSP auction, wherein the hybrid, optimized exchange is operably connected via a DSP to an advertiser having an advertisement available for purchase, wherein the hybrid, optimized exchange forwards to a DSP an advertising request that the hybrid, optimized exchange determines is likely to be relevant to the DSP, wherein the hybrid, optimized exchange bid maximizes a hybrid, optimized exchange profit that the hybrid, optimized exchange realizes in a winning SSP auction, wherein the hybrid, optimized exchange is further configured simultaneously to optimize a shading factor and a bid floor, wherein the hybrid, optimized exchange bid comprises the winning DSP bid divided by the optimized exchange shading factor, further comprising a learning engine configured to determine one or more of the optimized exchange bid floor and the optimized exchange shading factor, wherein the learning engine determines one or more of the optimized exchange bid floor and the optimized exchange shading factor once every training interval, wherein the training interval comprises a period of time after a conclusion of which the hybrid, optimized exchange, using the learning engine, performs a periodic updated calculation of one or more of the optimal bid floor and the optimal shading factor, wherein after determining the one or more of the optimized exchange bid floor and the optimized exchange shading factor, the learning engine resets a range of exploration around both the optimized exchange bid floor and the optimized exchange shading factor,
wherein the hybrid, optimized exchange further comprises an exchange controller, the exchange controller configured, using the bid floor, to conduct the hybrid, optimized exchange DSP auction,
wherein the hybrid, optimized exchange is further configured to determine the winning DSP bid and the winning DSP,
wherein the exchange controller comprises an exploration controller configured to receive the advertising request, wherein the exploration controller is further configured randomly to assign the received advertising request into one of a plurality of advertising request groups,
wherein the exploration controller is further configured randomly to assign the received advertising request into one of three advertising request groups, 1) a baseline group comprising a default set of baseline advertising requests using one or more of a default bid floor and a default shading factor whose performance the exploration controller can compare to a performance of other advertising requests using one or more of the optimized exchange bid floor and the optimized exchange shading factor, 2) an exploration group comprising exploration advertising requests, the exploration group usable by the exploration controller to gather a useful set of training data regarding the advertising requests, and 3) an exploitation group comprising exploitation advertising requests, the exploitation group usable by the exploration controller to exploit one or more of the optimized exchange bid floor and the optimized exchange shading factor,
wherein the exchange controller further comprises a bid parameter controller operably connected to the exploration controller, the bid parameter controller configured to determine one or more of an optimized exchange bid floor usable in the hybrid, optimized exchange DSP auction and an optimized exchange shading factor usable in the SSP auction, wherein the bid parameter controller obtains the one or more of the optimized exchange bid floor and the optimized exchange shading factor by making a bid parameter query, wherein the bid parameter query comprises a query for both the optimized exchange bid floor and the optimized exchange shading factor for a given advertising request source, a given advertising request country, and a given SSP auction type, wherein the bid parameter controller receives the advertising request groups from the exploration controller, wherein, using the advertising request groups, the bid parameter controller determines both the optimized exchange bid floor and the optimized exchange shading factor,
further comprising a database, the database operably connected to the exchange controller, wherein the learning engine is operably connected to the database,
wherein the hybrid, optimized exchange calculates the optimized exchange bid floor and the optimized exchange shading factor for a given subset of advertising requests, summing over all the advertising requests comprised in the subset, using the equation:
α*,β*=argmax Σprofit(ν,α,β)/|Bα,β| (1),
where the summation is taken over all the requests having the given advertising request source, the given advertising request country, and the given SSP auction type, and where:
α*, β* are calculated optimized exchange values for bid floor and shading factor, respectively,
ν is an application identifier,
α, β are values for bid floor and shading factor, respectively, in the training data, and
|Bα,β| is a size of a bucket comprising α and β.
|