Methods and apparatus to determine synthetic respondent level data using constrained Markov chains

Patent metadata

Field	Value
Publication number	US-10382818-B2
Application number	US-201715635153-A
Country	US
Kind code	B2
Filing date	Jun 27, 2017
Priority date	Jun 27, 2017
Publication date	Aug 13, 2019
Grant date	Aug 13, 2019

Abstract

Methods, apparatus, systems, and articles of manufacture are disclosed to generate synthetic respondent level data. An example method includes generating a first and second synthetic panel corresponding to a first duration of time based on a seed panel and first duration constraints; generating synthetic respondent level data corresponding to a second duration of time by linking synthetic panelists from the first synthetic panel to the second synthetic panel; and generating an output file based on the synthetic respondent level data.

First claim

What is claimed is:

1. An apparatus comprising:
a daily synthetic panel generator to generate, by executing an instruction with a processor, a first synthetic panel corresponding to a first duration of time and a second synthetic panel corresponding to a second duration of time, the first synthetic panel being based on (a) a seed panel and (b) first duration constraints based on return path data reported by a plurality of media devices, the second synthetic panel being based on (a) the seed panel and (b) second duration constraints based on the return path data reported by the plurality of media devices, the first synthetic panel representative of audiences of media presented by the plurality of media devices during the first duration of time, the second synthetic panel representative of audiences of media presented by the plurality of media devices during the second duration of time;
an extended synthetic panel generator to generate, by executing an instruction with the processor, synthetic respondent level data corresponding to a third duration of time by linking first synthetic panelists in the first synthetic panel to second synthetic panelist in the second synthetic panel; and
an output file generator to generate an output file based on the synthetic respondent level data.

2. The apparatus of claim 1 , wherein the daily synthetic panel generator is to generate the first synthetic panel by:
generating a transition matrix corresponding to the first and second durations of time, the transition matrix being based on the seed panel;
generating viewing data for the first synthetic panelists corresponding to the first and second durations of time, the viewing data being generated using the transition matrix;
determining weights to apply to the viewing data for the first synthetic panelists to satisfy the first duration constraints; and
removing one or more of the first synthetic panelists whose weights do not satisfy a minimum threshold.

3. The apparatus of claim 2 , wherein the transition matrix corresponds to a probability that a panelist will change stations during a quarter hour.

4. The apparatus of claim 2 , wherein the first synthetic panelists are reweighted after the one or more of the first synthetic panelists have been removed.

5. The apparatus of claim 1 , wherein the first duration constraints include reaches for the first duration corresponding to aggregate reach data determined based on the return path data.

6. The apparatus of claim 5 , wherein the extended synthetic panel generator is to generate the synthetic respondent level data corresponding to the third duration of time by:
determining a duration constraint error corresponding to a comparison of a first reach for the second duration determined by combining viewing data for synthetic panelists according to the synthetic panelist links and a second reach for the second duration corresponding to aggregate reach data determined based on the return path data; and
when the duration constraint error does not satisfy an error threshold, swapping the synthetic panelist links to reduce the duration constraint error.

7. The apparatus of claim 6 , wherein the swapping of the synthetic panelist links includes:
identifying the duration constraint error for permutations of panelist links between the first synthetic panel and the second synthetic panel; and
swapping the synthetic panelist links based on a highest reduction of the duration constraint error.

8. The apparatus of claim 1 , wherein the synthetic panelists are linked according to a demographic.

9. The apparatus of claim 1 , wherein the first duration of time corresponds to a first day, the second duration of time corresponds to a second day, and the third duration of time corresponds to a duration of time including the first and second days.

10. A method comprising:
generating, by executing an instruction with a processor, a first synthetic panel corresponding to a first duration of time and a second synthetic panel corresponding to a second duration of time, the first synthetic panel being based on a (a) seed panel and (b) first duration constraints based on return path data reported by a plurality of media devices, the second synthetic panel being based on (a) the seed panel and (b) second duration constraints based on the return path data reported by the plurality of media devices, the first synthetic panel representative of audiences of media presented by the plurality of media devices during the first duration of time, the second synthetic panel representative of audiences of media presented by the plurality of media devices during the second duration of time;
generating, by executing an instruction with the processor, synthetic respondent level data corresponding to a third duration of time by linking first synthetic panelists in the first synthetic panel to second synthetic panelist in the second synthetic panel; and
generating an output file based on the synthetic respondent level data.

11. The method of claim 10 , wherein the generating of the first synthetic panel includes:
generating a transition matrix corresponding to the first and second durations of time, the transition matrix being based on the seed panel;
generating viewing data for the first synthetic panelists corresponding to the first and second durations of time, the viewing data being based on the transition matrix;
determining weights to apply to the viewing data for the first synthetic panelists to satisfy the first duration constraints; and
removing one or more of the first synthetic panelists whose weights do not satisfy a minimum threshold.

12. The method of claim 11 , wherein the transition matrix corresponds to a probability that a panelist will change stations during a quarter hour.

13. The method of claim 11 , wherein the first synthetic panelists are reweighted after the one or more of the first synthetic panelists have been removed.

14. The method of claim 10 , wherein the first duration constraints include reaches for the first duration corresponding to aggregate reach data determined based on the return path data.

15. The method of claim 14 , wherein the generating of the synthetic respondent level data corresponding to the third duration of time includes:
determining a duration constraint error corresponding to a comparison of a first reach for the second duration determined by combining viewing data for synthetic panelists according to the synthetic panelist links and a second reach for the second duration corresponding to aggregate reach data determined based on the return path data; and
when the duration constraint error does not satisfy an error threshold, swapping the synthetic panelist links to reduce the duration constraint error.

16. The method of claim 15 , wherein the swapping of the synthetic panelist links includes:
identifying the duration constraint error for permutations of panelist links between the first synthetic panel and the second synthetic panel; and
swapping the synthetic panelist links based on a highest reduction of the duration constraint error.

17. The method of claim 10 , wherein the synthetic panelists are linked according to a demographic.

18. The method of claim 10 , wherein the first duration of time corresponds to a first day, the second duration of time corresponds to a second day, and the third duration of time corresponds to a duration of time including the first and second days.

19. A tangible computer readable storage medium comprising instructions which, when executed, cause a

Assignees

• Nielsen Co Us Llc
• The Nielson Company Us Llc

Inventors

• Sheppard Michael
• Sullivan Jonathan
• Morgan Michael D
• Shankar Balachander
• Murphy Edward
• Downing Frank

Classifications

• H04N21/25883

  being end-user demographical data, e.g. age, family status or address (arrangements for identifying locations of users in broadcast systems H04H60/52) · CPC title

• H04H60/31

  Arrangements for monitoring the use made of the broadcast services · CPC title

• H04N21/812

  involving advertisement data (advertising per se G06Q30/02) · CPC title

• H04N21/458

  Scheduling content for creating a personalised stream, e.g. by combining a locally stored advertisement with an incoming stream; Updating operations, e.g. for OS modules {; time-related management operations (arrangements for replacing or switching information during the broadcast or during the distribution H04H20/10)} · CPC title

• H04N21/44213 Primary

  Monitoring of end-user related data (arrangements for monitoring the users' behaviour or opinions in broadcast systems H04H60/33) · CPC title