Systems and methods for determining bids for placing advertisements

1. A computer-implemented method for determining a bid price of an advertisement, the method comprising:

dividing a time period beginning at an initial time into a set of learning periods associated with the advertisement;

generating a set of expected revenue rates for the set of learning periods;

estimating, using a processor, a cost for communicating the advertisement over the Internet, in the learning period starting at the initial time;

determining, using a processor, a current option value based on the estimated cost by:

(a) for a final learning period in the set of learning periods, determining a continuation value representing the expected profit over the final learning period and an option value representing a value of the option to stop communicating the advertisement over the Internet, at the beginning of the final learning period;

(b) starting with a learning period immediately preceding the final learning period,

(i) determining a continuation value representing a value of continuing to communicate the advertisement over the Internet, the continuation value based on the expected revenue rate for the learning period and the continuation values of later learning periods,

(ii) determining an option value representing a value of the option to stop communicating the advertisement over the Internet, at the beginning of the learning period, the option value based on the continuation value for the learning period, and

(iii) repeating steps (i) and (ii) for each prior learning period until an option value is determined for the learning period starting at the initial time; and

(c) a setting the current option value equal to the option value determined for the learning period starting at the initial time;

if the current option value is not zero, adjusting the estimated cost and repeating steps (a), (b), and (c);

if the current option value is zero, setting the bid price equal to the estimated cost; and

communicating, over the Internet, the bid price to an advertising exchange that communicates online advertising over the Internet.

2. The computer-implemented method of claim 1 , wherein generating the set of expected revenue rates further comprises:

associating a set of expected revenue rates with the learning period starting at the initial time;

assigning a set of transition probabilities to an initial expected revenue rate in the set of expected revenue rates, wherein the set of transition probabilities defines a relationship between the initial expected revenue rate and one or more later expected revenue rates for a second learning period; and

determining the initial expected revenue based on at least one of the transition probabilities and at least one of the later expected revenue rates.

3. The computer-implemented method of claim 1 , wherein the duration of each learning period is calculated as a function, at least in part, of a measurement noise and an initial revenue-rate variance.

4. The computer-implemented method of claim 3 , wherein the duration of each learning period is further calculated as a function, at least in part, of a variance-reducing parameter that determines a reduction in the initial revenue-rate variance.

5. The computer-implemented method of claim 4 , wherein the variance-reducing parameter is defined by a user.

6. The computer-implemented method of claim 1 , further including creating a lattice structure that includes a set of nodes, wherein an initial node corresponds to the initial expected revenue in the set of expected revenue rates.

7. The computer-implemented method of claim 6 , wherein the location of the initial node in the lattice structure is a function, at least in part, of the initial expected revenue rate.

8. The computer-implemented method of claim 6 , wherein the set of transition probabilities associated with the initial expected revenue rate is further associated with the corresponding initial node.

9. The computer-implemented method of claim 8 , wherein the set of transition probabilities contains

a higher-revenue transition probability defining a relationship between the initial node and a first later node in a second learning period,

a neutral-revenue transition probability defining a relationship between the initial node and a second later node in the second learning period, and

a lower-revenue transition probability defining a relationship between the initial node and a third later node in the second learning period.

10. The computer-implemented method of claim 9 , wherein the difference in revenue between the first later node and second later node equals the difference in revenue between the second later node and the third later node.

11. The computer-implemented method of claim 9 , wherein the higher-revenue transition probability equals the lower-revenue transition probability.

12. The computer-implemented method of claim 1 , further including:

calculating a learning value as a function, at least in part, of the bid price and the expected revenue rate of the initial learning period.

13. A computer-implemented method for determining a bid price of an advertisement, the method comprising:

dividing a time period beginning at an initial time into a set of learning periods associated with the advertisement;

generating a set of expected revenue rates for each learning period in the set of learning periods;

estimating, using a processor, a cost for communicating the advertisement over the Internet;

determining, using a processor, a current option value for the initial learning period based on the estimated cost by recursively determining, for each learning period beginning with a final learning period in the set of learning periods:

a continuation value associated with the expected revenue rates for the learning period and the continuation values of later learning periods, and

an option value, based on the continuation value, associated with an option to stop communication of the advertisement over the Internet during the learning period;

until the current option is zero, adjusting the estimated cost and repeating the determining step;

when the current option value is zero, setting the bid price equal to the estimated cost; and

communicating, over the Internet, the bid price to an advertising exchange that communicates online advertising over the Internet.

14. The computer-implemented method of claim 13 , wherein generating expected revenue rates further comprises:

associating a set of expected revenue rates with the learning period starting at the initial time;

assigning a set of transition probabilities to an initial expected revenue rate in the set of expected revenue rates, wherein the set of transition probabilities defines a relationship between the initial expected revenue rate and one or more later expected revenue rates for a second learning period; and

determining the initial expected revenue based on at least one of the transition probabilities and at least one of the later expected revenue rates.

15. The computer-implemented method of claim 13 , wherein the duration of each learning period is calculated as a function, at least in part, of a measurement noise and an initial revenue-rate variance.

16. The computer-implemented method of claim 15 , wherein the duration of each learning period is further calculated as a function, at least in part, of a variance-reducing parameter that determines a reduction in the initial revenue-rate variance.

17. The computer-implemented method of claim 16 , wherein the variance-reducing parameter is defined by a user.

18. The computer-implemented method of claim 13 , further including creating a lattice structure that includes a set of nodes, wherein an initial node corresponds to the initial expected revenue in the set of expected revenue rates.

19. The computer-implemented method of claim 18 , wherein the location of the initial node in the lattice structure is a function, at least in part, of the initial expected revenue rate.

20. The computer-implemented method of claim 18 , wherein the set of transition probabilities associated with the initial expected revenue rate is further associated with the corresponding initial node.

21. The computer-implemented method of claim 20 , wherein the set of transition probabilities contains a higher-revenue transition probability defining a relationship between the initial node and a first later node in a second learning period, a neutral-revenue transition probability defining a relationship between the initial node and a second later node in the second learning period, and a lower-revenue transition probability defining a relationship between the initial node and a third later node in the second learning period.

22. The computer-implemented method of claim 21 , wherein the difference in revenue between the first later node and second later node equals the difference in revenue between the second later node and the third later node.

23. The computer-implemented method of claim 21 , wherein the higher-revenue transition probability equals the lower-revenue transition probability.

24. The computer-implemented method of claim 13 , further including:

calculating a learning value as a function of at least the bid price and the expected revenue rate of the initial learning period.

25. A bid calculating apparatus for determining a bid price of an advertisement, the bid calculating apparatus comprising:

a storage device storing instructions for determining a bid price of an advertisement; and

a processing device configured to execute the instructions such that the processing device is configured to operate:

a learning period module configured to divide a time period beginning at an initial time into a set of learning periods associated with the advertisement, and estimate a cost for communicating the advertisement over the Internet, in the learning period starting at the initial time;

a lattice module configured to

generate a set of expected revenue rates for each learning period in the set of learning periods; and

determine a current option value for the initial learning period based on the estimated cost by recursively determining, for each learning period beginning with a final learning period in the set of learning periods:

a continuation value associated with the expected revenue rate for the learning period and the continuation values of later learning periods, and

an option value, based on the continuation value, associated with an option to stop communicating the advertisement over the Internet, in the learning period; and

a bid generator configured to

determine whether the current option value, when set to equal the option value of the learning period starting at the initial time, is zero;

adjust the estimated cost and direct the lattice module to repeat the determining function, when it is determined that the current option value is not zero; and

set the bid price equal to the estimated cost, when it is determined that the current option value is zero.