Techniques for displaying impressions in documents delivered over a computer network

1. A method comprising:

receiving a request for an impression to be displayed in a placement of a web page;

receiving a placement identifier particularly identifying the placement for the impression in the web page;

receiving behavioral data indicative of an online history and previously received impressions;

determining a plurality of candidate impressions that may be displayed in the placement of the web page, based, at least in part, on the behavioral data;

determining a learning mode of each candidate impression, the learning mode of the each candidate impression being indicative of a number of times the each candidate impression has been served,

wherein the learning mode is one of a plurality of learning modes, the plurality of learning modes comprising a first learning mode, a second learning mode, and a third learning mode,

wherein the first learning mode includes candidate impressions that have been served less than or equal to a first predetermined threshold number of times,

wherein the second learning mode includes candidate impressions that have been served more than the first predetermined threshold number of times and less than or equal to a second predetermined threshold number of times, and

wherein the third learning mode includes candidate impressions that have been served more than the second predetermined threshold number of times;

applying learning mode percentages to determine a retained learning mode from the plurality of learning modes; and

selecting a selected impression from the plurality of candidate impressions that are in the retained learning mode.

2. The method of claim 1 , wherein determining the plurality of candidate impressions comprises:

using at least the behavioral data and the placement identifier to determine a first set of impressions that may be displayed in the web page;

generating a second set of impressions from the first set of impressions by applying business rules to the first set of impressions, the business rules being designed to increase a likelihood of interest in impressions in the second set of impressions.

3. The method of claim 2 , wherein the business rules include a rule that impressions previously displayed are not included in the second set of impressions.

4. The method of claim 1 , wherein the placement identifier comprises a hypertext markup language (HTML) tag.

5. The method of claim 1 , wherein each impression in the plurality of candidate impressions comprises a message and an associated creative, the creative comprising a design of how the message is to be presented.

6. The method of claim 5 , wherein the plurality of candidate impressions comprises:

a first message with an associated first creative;

the first message with an associated second creative; and

the first message with an associated third creative;

wherein the first, second, and third creatives are different from each other.

7. The method of claim 1 , wherein:

the first learning mode is a non-scaled learning mode;

the second learning mode is a globally scaled learning mode; and

the third learning mode is a fully scaled learning mode.

8. The method of claim 1 , wherein the behavioral data are received in a cookie.

9. The method of claim 1 , wherein the placement identifier is received as part of the request.

10. The method of claim 1 , wherein selecting the selected impression comprises:

randomly selecting the selected impression when all impressions in the retained learning mode are in the first learning mode; and

selecting a highest revenue generating impression when all impressions in the retained learning mode are in the second learning mode or the third learning mode.

11. An impressions server configured to:

receive a request for an impression to be displayed in a placement of a web page;

receive a placement identifier identifying the placement of the impression in the web page, the placement identifier embedded in the web page;

receive behavioral data comprising information indicative of actions taken on previously visited websites and information indicative of impressions previously displayed;

select a set of candidate impressions from a set of impressions using at least the behavioral data and the placement identifier;

wherein each impression in the set of candidate impressions has an associated learning mode, the learning mode of each the impression being indicative of a number of times the each impression has been served,

wherein the learning mode is one of a plurality of learning modes, the plurality of learning modes comprising a first learning mode, a second learning mode, and a third learning mode,

wherein the first learning mode includes impressions that have been served less than or equal to a first predetermined threshold number of times,

wherein the second learning mode includes impressions that have been served more than the first predetermined threshold number of times and less than or equal to a second predetermined threshold number of times, and

wherein the third learning mode includes impressions that have been served more than the second predetermined threshold number of times;

select a selected impression based at least in part on a learning mode of each impression in the set of candidate impressions; and

forward the selected impression to display in the placement of the web page.

12. The impressions server of claim 11 , wherein the behavioral data are stored in a cookie received by the impressions server.

13. The impressions server of claim 12 , further comprising a behavior monitor including computer-readable program code for monitoring online-activities, the behavior monitor being configured to facilitate storage in the cookie categories of previously visited websites.

14. A method comprising:

receiving a request for an impression to be displayed in a placement of a web page;

receiving a cookie comprising data including behavioral data indicative of categories of websites previously visited; and

using at least the behavioral data to select a selected impression from a plurality of candidate impressions that may be displayed in the placement of the web page, the selected impression being selected from the plurality of candidate impressions based at least in part on anticipated revenue generation,

wherein selecting the selected impression from the plurality of candidate impressions comprises:

randomly selecting the selected impression from the plurality of candidate impressions when all impressions in the plurality of candidate impressions are in a first learning mode, and

selecting a highest revenue generating impression in the plurality of candidate impressions as the selected impression when all impressions in the plurality of candidate impressions are in a second learning mode, and

selecting a highest revenue generating impression in the plurality of candidate impressions as the selected impression when all of the impressions in the plurality of candidate impressions are in a third learning mode;

wherein the first learning mode includes impressions that have been served less than or equal to a first predetermined threshold number of times,

wherein the second learning mode includes impressions that have been served more than the first predetermined threshold number of times and less than or equal to a second predetermined threshold number of times, and

wherein the third learning mode includes impressions that have been served more than the second predetermined threshold number of times.

15. The method of claim 14 , further comprising:

receiving a placement identifier identifying the placement.

16. The method of claim 14 , wherein the plurality of candidate impressions comprise:

a first message with an associated first creative;

the first message with an associated second creative; and

the first message with an associated third creative;

wherein the first, second, and third creatives being different from each other.

17. The method of claim 14 , wherein selecting the selected impression from the plurality of candidate impressions further comprises:

selecting impressions in the plurality of candidate impressions that are all in the first learning mode, the second learning mode, or in the third learning mode when at least some impressions in the plurality of candidate impressions are in different learning modes.

18. The method of claim 14 , wherein:

the first learning mode is a non-scaled learning mode,

the second learning mode is a globally scaled learning mode, and

the third learning mode is a fully scaled learning mode.

19. A method, comprising:

receiving a request for an impression to be displayed in a placement of a web page

receiving a placement identifier particularly identifying the placement for the impression in the web page;

receiving behavioral data indicative of an online history and previously received impressions;

determining a plurality of candidate impressions that may be displayed in the placement of the web page, based, at least in part, on the behavioral data;

determining a learning mode of each candidate impression, the learning mode of the each candidate impression being indicative of a number of times the each candidate impression has been served,

wherein the learning mode is one of a plurality of learning modes, the plurality of learning modes comprising a first learning mode, a second learning mode, and a third learning mode,

wherein the first learning mode includes candidate impressions that have been served less than or equal to a first predetermined threshold number of times,

wherein the second learning mode includes candidate impressions that have been served more than the first predetermined threshold number of times and less than or equal to a second predetermined threshold number of times, and

wherein the third learning mode includes candidate impressions that have been served more than the second predetermined threshold number of times; and

selecting a selected impression from the plurality of candidate impressions.

20. The method of claim 19 , wherein:

the first learning mode is a non-scaled learning mode,

the second learning mode is a globally scaled learning mode, and

the third learning mode is a fully scaled learning mode.