INFERRING LEGITIMACY OF ADVERTISEMENT CALLS

• US 20090012853A1
• Filed: 07/03/2007
• Published: 01/08/2009
• Est. Priority Date: 07/03/2007
• Status: Abandoned Application

• Current Assignee: Oath Inc. (Verizon Communications Inc.)
• Original Assignee: Right Media Incorporated (Apollo Global Management, Inc.)
• Inventors: Dalto, John H., Nolet, Michiel
• Application Number:
  US11/772,979
• Publication Number:
  US 20090012853A1
• Time in Patent Office: Days
• US Class Current: 705/14.470
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0248 Avoiding fraud
• CPC Class Codes: G06Q 30/0249 based upon budgets or funds
• CPC Class Codes: G06Q 30/0273 Determination of fees for advertising
• CPC Class Codes: G06Q 30/0277 Online advertisement

Abstract

There are methods and apparatus, including computer program products, for receiving advertisement calls at a first computing system from a second computing system, the first computing system and a second computing system being in electronic communication through a network, each advertisement call being defined by one or more variable-value pairs; extracting data from the advertisement calls, the extracted data including at least two sets of variable/value pairs, the first set of variable/value pairs including variable/value pairs of a first variable type, and the second set of variable/value pairs including variable/value pairs of a second variable type; and performing one or more tests on the extracted data to infer a legitimacy of at least a first subset of the advertisement calls.

Claims

1. A computer-implemented method comprising:

receiving advertisement calls at a first computing system from a second computing system, the first computing system and a second computing system being in electronic communication through a network, each advertisement call being defined by one or more variable-value pairs;

extracting data from the advertisement calls, the extracted data including at least two sets of variable/value pairs, the first set of variable/value pairs including variable/value pairs of a first variable type, and the second set of variable/value pairs including variable/value pairs of a second variable type; and

performing one or more tests on the extracted data to infer a legitimacy of at least a first subset of the advertisement calls.

2. The method of claim 1, wherein the first set of variable/value pairs consists of variable/value pairs of the first variable type.

3. The method of claim 1, wherein the second set of variable/value pairs consists of variable/value pairs of the second variable type.

4. The method of claim 1, wherein at least the first subset of the advertisement calls are received from a first user agent of the second computing system.

5. The method of claim 4, wherein the first user agent is operable by a human user.

6. The method of claim 4, wherein the first user agent is operable by a robot.

7. The method of claim 4, wherein at least a second subset of the advertisement calls are received from a second user agent of the second computing system.

8. The method of claim 4, wherein at least a second subset of the advertisement calls are received from a user agent of a third computing system.

9. The method of claim 1, wherein the first variable type comprises an impression frequency and the second variable type comprises an impression recency.

10. The method of claim 1, wherein performing one or more tests on the extracted data comprises:

determining a distribution of impressions over impression frequency and impression recency.

11. The method of claim 10, further comprising:

taking an action based on the determined distribution of impressions over impression frequency and impression recency.

12. The method of claim 11, wherein taking an action comprises:

flagging at least the first subset of the advertisement calls as being associated with fraudulent activity on an advertisement exchange or an advertisement network if the distribution of impressions over impression frequency and impression recency is determined to satisfy one or more conditions.

13. The method of claim 11, wherein taking an action comprises:

flagging at least the first subset of the advertisement calls as being associated with fraudulent activity on an advertisement exchange or an advertisement network if the distribution of impressions over impression frequency and impression recency is determined to be skewed toward one extremum of a distribution spectrum.

14. The method of claim 11, wherein taking an action comprises:

identifying a slice of inventory associated with the first subset of the advertisement calls as being associated with fraudulent activity on an advertisement exchange or an advertisement network based on the determined distribution; and

suspending the identified slice of inventory from being transacted on the advertisement exchange or the advertisement network.

15. The method of claim 14, wherein suspending the identified slice of inventory comprises:

suspending non-cost-per-action-based items of inventory within the identified slice of inventory.

16. The method of claim 14, wherein suspending the identified slice of inventory comprises:

deactivating the identified slice of inventory.

17. The method of claim 1, wherein the first variable type comprises a number of impressions, and a second variable type comprises a number of clicks.

18. The method of claim 17, further comprising:

calculating click rates for a slice of inventory associated with the first subset of the advertisement calls based on the values of the first set of variable/value pairs and the values of the second set of variable/value pairs.

19. The method of claim 18, wherein the extracted data further comprises a third set of variable/value pairs including variable/value pairs of a third variable type.

20. The method of claim 19, wherein the third variable type comprises an impression frequency, and wherein performing one or more tests on the extracted data comprises:

performing an autocorrelation of variables test to determine whether there is a correlation between clicks rates and impression frequency.

21. The method of claim 19, wherein the third variable type comprises an impression recency, and wherein performing one or more tests on the extracted data comprises:

performing an autocorrelation of variables test to determine whether there is a correlation between clicks rates and impression recency.

22. The method of claim 19, wherein the third variable type comprises a uniform resource locator (URL) frequency, and wherein performing one or more tests on the extracted data comprises:

performing an autocorrelation of variables test to determine whether there is a correlation between clicks rates and URL frequency.

23. The method of claim 19, wherein the third variable type comprises an advertisement type, and wherein a value assigned to the advertisement type comprises one of the following:

a value indicative of a Flash-type advertisement and a value indicative of a GIF-type advertisement.

24. The method of claim 1, wherein performing one or more tests on the extracted data comprises:

taking an action if at least two of the following conditions are satisfied;

(a) a number of impressions is greater than a predefined threshold, (b) click rate associated with Flash-type advertisements is zero; and

(c) click rate associated with GIF-type advertisements is zero.

25. The method of claim 24, wherein taking an action comprises:

flagging at least the first subset of the advertisement calls as being associated with suspicious activity on an advertisement exchange or an advertisement network.

26. The method of claim 24, wherein taking an action comprises:

identifying the slice of inventory as being associated with suspicious activity on an advertisement exchange or an advertisement network.

27. The method of claim 1, wherein performing one or more tests on the extracted data comprises:

performing an autocorrelation of variables tests to determine a degree of correlation between clicks rates and one or more of the following;

impression frequency, impression recency, and uniform resource locator frequency; and

taking an action based on one or more of the determined degrees of correlation.

28. The method of claim 27, wherein taking an action comprises:

flagging at least the first subset of the advertisement calls as being associated with suspicious activity on an advertisement exchange or an advertisement network if one or more of the determined degrees of correlation satisfies one or more conditions.

29. The method of claim 27, wherein taking an action comprises:

identifying a slice of inventory associated with the first subset of the advertisement calls as being associated with suspicious activity on an advertisement exchange or an advertisement network based on one or more of the determined degrees of correlation.

30. The method of claim 1, wherein performing one or more tests on the extracted data comprises:

performing a conditional probabilities test to determine whether a slice of inventory is performing at an extremum of a spectrum with respect to conversions.

31. The method of claim 1, wherein performing one or more tests on the extracted data comprises:

performing an autocorrelation of variables tests to determine a degree of correlation between clicks rates and one or more of the following;

impression frequency, impression recency, and uniform resource locator frequency;

performing a conditional probabilities test to determine whether a slice of inventory is performing at an extremum of a spectrum with respect to conversions; and

performing an advertisement type test in which click rates associated with GIF-type advertisements and Flash-type advertisements are examined.

32. The method of claim 31, further comprising:

flagging at least the first subset of the advertisement calls as being associated with suspicious activity on an advertisement exchange or an advertisement network based on the results of any one of the tests.

33. The method of claim 31, further comprising:

flagging at least the first subset of the advertisement calls as being associated with fraudulent activity on the advertisement exchange or the advertisement network based on the results of at least two of the tests.

34. The method of claim 31, further comprising:

suspending a slice of inventory associated with the first subset of the advertisement calls from being transacted on the advertisement exchange or the advertisement network if the first set of advertisement calls is flagged as being associated with fraudulent activity on the advertisement exchange or the advertisement network.

35. The method of claim 1, further comprising:

suspending a slice of inventory associated with the first subset of the advertisement calls from being transacted on the advertisement exchange or the advertisement network if at least two of the tests that are performed indicate that the first set of advertisement calls is associated with fraudulent activity on the advertisement exchange or the advertisement network.

36. The method of claim 1, further comprising:

suspending non-cost-per-action-based items of inventory within a slice of inventory associated with the first subset of the advertisement calls from being transacted on the advertisement exchange or the advertisement network if at least two of the tests that are performed indicate that the first set of advertisement calls is associated with fraudulent activity on the advertisement exchange or the advertisement network.

37. A machine-readable medium that stores executable instructions to cause a machine to:

receive advertisement calls at a first computing system from a second computing system, the first computing system and a second computing system being in electronic communication through a network, each advertisement call being defined by one or more variable-value pairs;

extract data from the advertisement calls, the extracted data including at least two sets of variable/value pairs, the first set of variable/value pairs including variable/value pairs of a first variable type, and the second set of variable/value pairs including variable/value pairs of a second variable type; and

perform one or more tests on the extracted data to infer a legitimacy of at least a first subset of the advertisement calls.

38. The machine-readable medium of claim 37, wherein instructions to cause the machine to perform one or more tests comprise instructions to:

determine a distribution of impressions over impression frequency and impression recency.

39. The machine-readable medium of claim 37, wherein instructions to cause the machine to perform one or more tests comprise instructions to:

perform an autocorrelation of variables test to determine whether there is a correlation between clicks rates and impression frequency.

40. The machine-readable medium of claim 37, wherein instructions to cause the machine to perform one or more tests comprise instructions to:

perform an autocorrelation of variables test to determine whether there is a correlation between clicks rates and impression recency.

41. The machine-readable medium of claim 37, wherein instructions to cause the machine to perform one or more tests comprise instructions to:

perform an autocorrelation of variables test to determine whether there is a correlation between clicks rates and URL frequency.

42. The machine-readable medium of claim 37, wherein instructions to cause the machine to perform one or more tests comprise instructions to:

perform a conditional probabilities test to determine whether a slice of inventory is performing at an extremum of a spectrum with respect to conversions.

43. The machine-readable medium of claim 37, further comprising instructions to cause the machine to:

based on results of the one or more tests, flagging at least the first subset of the advertisement calls as being associated with suspicious activity on an advertisement exchange or an advertisement network.

44. The machine-readable medium of claim 37, further comprising instructions to cause the machine to:

based on results of the one or more tests, identify a slice of inventory associated with the first subset of the advertisement calls as being associated with suspicious activity on an advertisement exchange or an advertisement network; and

suspend the identified slice of inventory from being transacted on the advertisement exchange or the advertisement network.

45. The machine-readable medium of claim 44, wherein instructions to suspend the identified slice of inventory comprise instructions to:

suspend non-cost-per-action-based items of inventory within the identified slice of inventory.

46. A computer-implemented method comprising:

receiving advertisement calls for a slice of inventory on an advertisement exchange or an advertisement network, the advertisement call being received at a first computing system from a second computing system, the first computing system and a second computing system being in electronic communication through a network, each advertisement call being defined by one or more variable-value pairs;

extracting data from the advertisement calls, the extracted data including at least two sets of variable/value pairs, the first set of variable/value pairs including variable/value pairs of a first variable type, and the second set of variable/value pairs including variable/value pairs of a second variable type;

performing one or more tests on the extracted data to infer a legitimacy of at least a first subset of the advertisement calls;

identifying non-cost-per-action-based items of inventory within the slice that are associated with the first subset of the advertisement calls; and

based on the results of performing the one or more tests, suspending the identified non-cost-per-action-based items of inventory from being transacted on the advertisement exchange or an advertisement network.

47. A machine-readable medium that stores executable instructions to cause a machine to:

receive advertisement calls for a slice of inventory on an advertisement exchange or an advertisement network, the advertisement call being received at a first computing system from a second computing system, the first computing system and a second computing system being in electronic communication through a network, each advertisement call being defined by one or more variable-value pairs;

extract data from the advertisement calls, the extracted data including at least two sets of variable/value pairs, the first set of variable/value pairs including variable/value pairs of a first variable type, and the second set of variable/value pairs including variable/value pairs of a second variable type;

perform one or more tests on the extracted data to infer a legitimacy of at least a first subset of the advertisement calls;

identify non-cost-per-action-based items of inventory within the slice that are associated with the first subset of the advertisement calls; and

based on the results of the performance of the one or more tests, suspend the identified non-cost-per-action-based items of inventory from being transacted on the advertisement exchange or an advertisement network.