System and method using sampling for scheduling advertisements in slots of different quality in an online auction with budget and time constraints
An improved system and method is provided for using sampling for scheduling advertisements in slots of different quality in an online auction with budget and time constraints. A multi-armed bandit engine may be provided for sampling new advertisements by allocating advertisements for web page placements of different quality and optimizing payments to maximize the welfare of the advertisers while remaining within advertiser's budget and time constraints. Advertisers may report their private information including arrival time, departure time, value per click, and budget. And the multi-armed bandit mechanism may approximate the maximal welfare that may be achieved under budget and time constraints by bounding the possible gain from any possible lie an advertiser might submit in reporting private information. Advertisers departing from the online auction may be charged using a payment method that may provide truthful guarantees on budget, arrivals, departures, and valuations for a budget-constrained online auction.
1. A computer system for an online advertising auction, comprising:
a multi-armed bandit engine comprising:
an input/output interface for receiving from each of a plurality of advertisers bidding for advertising slots in the online advertising auction:
a bid for an advertising slot based on a keyword, wherein said bid comprises:
an arrival time at the online advertising auction;
a departure time at the online advertising auction;
a budget that may not be exceeded between the arrival time and the departure time, wherein said budget may not be exceeded; and
a value per click;
a processor device for:
selecting a subset of the plurality of the advertisers;
allocating a plurality of web page placements of different quality to the subset of the advertisers for sampling the advertisements in the online advertising auction, wherein quality is defined as a probability of a click-through if an advertisement appears in said web page placement, wherein said quality is known;
learning a valuation of the advertisements through sampling by scheduling the advertisements for web page placements of different quality in the online advertising auction to optimize payments for maximizing welfare of the advertisers;
updating a click-through rate using a normalized number of clicks for each of the subset of the plurality of advertisers sampled in the online advertising auction;
updating a payoff rate for receiving a click;
updating a remaining budget of each of the subset of the plurality of advertisers; and
charging the plurality of advertisers departing from the online advertising auction; and
a storage device operably coupled to the multi-armed bandit engine for storing a plurality of budgets and a plurality of bids each associated with an advertisement allocated to web page placements in the online advertising auction with budget and time constraints;
wherein the computer system selects a plurality of advertisers for sampling advertisements in the online advertising auction and samples advertisements for a subset of the plurality of advertisers.
2. The system of claim 1 further comprising a model generator for creating a multi-armed bandit model used by the multi-armed bandit engine.
3. The system of claim 1 further comprising a payoff optimizer operably coupled to the multi-armed bandit engine for optimizing payments for the advertisements sampled in the online advertising auction with budget and time constraints to maximize the welfare of the advertisers.
4. A method for an online advertising auction, the method comprising:
using an input/output interface for receiving from each of a plurality of advertisers bidding for advertising slots in the online advertising auction:
a budget that may not be exceeded between an arrival time and the departure time for each of a plurality of advertisers entering an online advertising auction; and
a value per click of an advertisement for each of the plurality of advertisers entering the online advertising auction;
using a processor device for:
selecting a subset of the plurality of advertisers;
allocating a plurality of web page placements of varying quality to the subset of the advertisers for sampling the advertisements in the online advertising auction, wherein quality is defined as a probability of a click-through if an advertisement appears in said web page placement;
sampling advertisements for the subset of the plurality of advertisers using the plurality of the web page placements of varying quality;
updating a click-through rate using a normalized number of clicks for each of the subset of the plurality of advertisers sampled in the online advertising auction;
updating a payoff rate for receiving a click for each of the subset of the plurality of advertisers sampled in the online advertising auction;
updating a remaining budget of each of the plurality of advertisers in the online advertising auction; and
charging the plurality of advertisers departing from the online advertising auction;
wherein each step of the method is computer-implemented.
5. The method of claim 4 further comprising receiving the arrival time and the departure time for each of the plurality of advertisers entering the online advertising auction.
6. The method of claim 4 further comprising initializing the click-through rate to zero for each of the plurality of advertisers entering the online advertising auction.
7. The method of claim 4 further comprising initializing total clicks of advertisements to zero for each of the plurality of web page placements for each of the plurality of advertisers entering the online advertising auction.
8. The method of claim 4 further comprising initializing the price charged for a click of an advertisement to zero for each of the plurality of advertisers entering the online advertising auction.
9. The method of claim 4 further comprising initializing an exposure parameter to zero for each of the plurality of advertisers entering the online advertising auction on a first visit.
10. The method of claim 4 wherein selecting the plurality of advertisers for sampling advertisements in the online advertising auction comprises:
determining a list of the plurality of advertisers currently in the online advertising auction;
removing advertisers with a click-through rate lower than a threshold from the list; and
outputting the remaining list of the plurality of advertisers for sampling advertisements in the online advertising auction.
11. The method of claim 4 wherein sampling advertisements for a subset of the plurality of advertisers in the online advertising auction comprises:
receiving a list of the plurality of advertisers for sampling advertisements in the online advertising auction;
randomly selecting a subset of advertisers from the list; and
sampling an advertisement for each of the subset of advertisers in the online advertising auction.
12. The method of claim 4 further comprising updating an exposure parameter for each of the subset of the plurality of advertisers sampled in the online advertising auction.
13. The method of claim 4 further comprising updating a click-through status set for each of the subset of the plurality of advertisers sampled in the online advertising auction that received a click of an advertisement.
14. The method of claim 4 further comprising computing critical values for each of the plurality of web page placements for each of the plurality of advertisers in the online advertising auction.
15. The method of claim 4 further comprising computing an interim price for each of the plurality of web page placements for each of the plurality of advertisers in the online advertising auction.
16. The method of claim 4 wherein updating a click-through rate using a normalized number of clicks for each of the subset of the plurality of advertisers sampled in the online advertising auction comprises updating the click-through rate for each of the subset of the plurality of advertisers sampled in the online advertising auction using probability constants.