Systems and methods for controlling online advertising campaigns

1. A method of controlling advertising campaigns comprising:

receiving, from an advertiser, a cost indicator that indicates an average cost per event to stay below for a current portion of an advertising campaign of the advertiser;

obtaining, by a processor via a cost sensor, real-time data associated with bid transactions of an advertising market associated with the advertising campaign;

monitoring, by the processor, one or more bid transactions of the advertising market in the real-time data;

determining, by the processor, an observed cost measurement from the one or more bid transactions based on the monitoring;

determining, by the processor, one or more cell-level performance predictions associated with the advertising market, wherein the one or more cell-level performance predictions comprise a first cell-level performance prediction of a first cell of the advertising market representing a likelihood that an impression delivered to the first cell will result in at least one of a click-through or a conversion and a second cell- level performance prediction of a second cell of the advertising market representing a likelihood that an impression delivered to the second cell will result in at least one of a click-through or a conversion;

determining, by the processor, an adaptive control scheme configured to calculate one or more signals based on a stored controller state variable comprising information about past control performance associated with one or more past bids;

receiving a pacing indicator that indicates a pacing to achieve for the current portion of the advertising campaign;

determining, by a campaign controller via the processor, a price control, using the adaptive control scheme, based on a function f(r, c, T cost max , B rev max ), wherein r represents an observed pacing measurement, c represents the observed cost measurement, B rev max represents the cost indicator, and B rev max represents the pacing indicator;

updating, by the processor, the price control to generate an updated price control, using the adaptive control scheme, based on a change to the pacing;

receiving, from the advertiser, a cost reference signal indicative of a cost restriction for the advertising campaign;

calculating, by the processor, a cost error based on the cost reference signal; and

controlling, by the campaign controller via the processor, a bid price for the advertising campaign based on the cost error and min(maxCPI, u p {circumflex over (p)} i ), wherein maxCPI represents max cost per impression, u p represents at least one of the price control or the updated price control, and {circumflex over (p)} i represents at least one of the one or more cell-level performance predictions, wherein the controlling is associated with optimizing performance of bidding in the advertising market in favor of the advertiser.

2. The method of claim 1 , comprising:

monitoring the bid transactions of the advertising market; and

determining the observed pacing measurement from the bid transactions.

3. The method of claim 2 , wherein determining the price control based on the observed pacing measurement comprises:

calculating a pacing error based on a difference between the pacing indicator and the observed pacing measurement; and

determining the price control based on the pacing error.

4. The method of claim 3 , wherein the adaptive control scheme is configured to calculate the one or more signals based on a most recent value of a cost error signal and a most recent value of a pacing error signal.

5. The method of claim 2 , wherein the price control is a single price control.

6. The method of claim 2 , wherein the pacing indicator is represented in monetary units.

7. The method of claim 2 , wherein the pacing indicator is represented as a number of events, wherein the number of events comprise at least one of impressions, clicks, or conversions.

8. The method of claim 2 , wherein the observed cost measurement is a moving average.

9. The method of claim 2 , wherein the observed pacing measurement is a moving average.

10. One or more non-transitory computer-readable media having instructions stored thereon, wherein the instructions, when executed by a processor of a computing device, cause the computing device to:

receive, from an advertiser, a cost indicator that indicates an average cost per event to stay below for a current portion of an advertising campaign of the advertiser;

receive, from the advertiser, a pacing indicator that indicates a pacing to achieve for the current portion of the advertising campaign;

obtain, by the processor via a cost sensor, real-time data associated with bid transactions of an advertising market associated with the advertising campaign;

monitor, by the processor, one or more bid transactions of the advertising market in the real-time data;

determine, by the processor, an observed cost measurement and an observed pacing measurement from the one or more bid transactions based on the monitoring;

determine, by the processor, one or more cell-level performance predictions associated with the advertising market, wherein the one or more cell-level performance predictions comprise a first cell-level performance prediction of a first cell of the advertising market representing a likelihood that an impression delivered to the first cell will result in at least one of a click-through or a conversion and a second cell-level performance prediction of a second cell of the advertising market representing a likelihood that an impression delivered to the second cell will result in at least one of a click-through or a conversion;

determine, by the processor, an adaptive control scheme configured to calculate one or more signals based on a stored controller state variable comprising information about past control performance;

receive, by the processor, a pacing indicator that indicates a pacing to achieve for the current portion of the advertising campaign;

determine, by a campaign controller via the processor, a price control, using the adaptive control scheme, based on a function f(r, c, T cost max , B rev max ), wherein r represents the observed pacing measurement, c represents, the observed cost measurement, the pacing indicator and the observed pacing measurement, T cost max represents the cost indicator, and B rev max represents the pacing indicator;

update, by the processor, the price control to generate an updated price control, using the adaptive control scheme, based on a change to the pacing; and

control, by the campaign controller via the processor, a bid price for the advertising campaign based on min(maxCPI, u p {circumflex over (p)} i ), wherein maxCPI represents max cost per impression, u p represents at least one of the price control or the updated price control and {circumflex over (p)} i represents at least one of the one or more cell-level performance predictions.

11. The one or more computer-readable media of claim 10 , wherein to determine the price control based on the cost indicator, the observed cost measurement, the pacing indicator, and the observed pacing measurement, the instructions cause the computing device to:

calculate of a cost error based on a difference between the cost indicator and the observed cost measurement;

calculate a pacing error based on a difference between the pacing indicator and the observed pacing measurement; and

determine the price control based on the cost error and the pacing error.

12. The one or more computer-readable media of claim 11 , wherein the adaptive control scheme is configured to calculate the one or more signals based on at least one of a most recent value of a cost error signal or a most recent value of a pacing error signal.

13. The one or more computer-readable media of claim 10 , wherein the price control is a single price control.

14. The one or more computer-readable media of claim 10 , wherein the pacing indicator is represented in monetary units.

15. The one or more computer-readable media of claim 10 , wherein the pacing indicator is represented as a number of events, wherein the number of events comprise at least one of impressions, clicks, or conversions.

16. A system for controlling advertising campaigns comprising:

one or more processors configured to implement:

a cost error calculator to:

receive a cost signal that indicates an average cost per event to stay below for a current portion of an advertising campaign and an observed cost signal based on observed transactions of the advertising campaign; and

generate a cost error signal based on the cost signal and the observed cost signal;

a pacing error calculator to:

receive a pacing signal that indicates a pacing to achieve for the current portion of the advertising campaign and an observed pacing signal based on one or more observed transactions of the advertising campaign; and

generate a pacing error signal based on the pacing signal and the observed pacing signal; and

a controller to:

determine one or more cell-level performance predictions associated with an advertising market associated with the advertising campaign, wherein the one or more cell-level performance predictions comprise a first cell-level performance prediction of a first cell of the advertising market representing a likelihood that an impression delivered to the first cell will result in at least one of a click-through or a conversion and a second cell-level performance prediction of a second cell of the advertising market representing a likelihood that an impression delivered to the second cell will result in at least one of a click-through or a conversion;

generate, via a campaign controller, a price control signal based on a function f(r, c, T cost max , B rev max ), wherein r represents an observed pacing measurement associated with the observed pacing signal, c represents an observed cost measurement associated with the observed cost signal, T cost max represents a cost indicator associated with the cost signal, and B rev max represents a pacing indicator associated with the pacing signal;

update the price control signal to generate an updated price control signal based on a change to the pacing; and

control, via the campaign controller, a bid price for the advertising campaign based on maxCPI, u p , and {circumflex over (p)} i , wherein maxCPI represents max cost per impression, u p represents at least one of the price control signal or the updated price control signal and {circumflex over (p)} i represents at least one of the one or more cell-level performance predictions, wherein the controlling is associated with optimizing performance of bidding in the advertising market in favor of an advertiser.

17. The system of claim 16 , wherein to generate the price control signal based on the cost error signal and the pacing error signal is accomplished utilizing an adaptive control scheme.

18. The system of claim 16 , comprising a cost actuator to:

calculate a bid in accordance with the price control signal.

19. The system of claim 18 , wherein the cost actuator is to receive one or more performance predictions for inventory cells of the advertising campaign and wherein to calculate a bid in accordance with the price control signal is to calculate a bid for each of the inventory cells in accordance with both the one or more performance predictions of the inventory cells and the price control signal.

20. The system of claim 18 , wherein the cost actuator is to place the bid in an advertising marketplace associated with the advertising campaign.