Methods and apparatus for session building from ping-level data
Methods, apparatus, systems and articles of manufacture to build viewing sessions from ping-level data are disclosed herein. An example apparatus includes a threshold calculator to determine a threshold based on reference sessions represented in panel data obtained from panel meters. The example apparatus also includes a session partitioner to determine if a first ping from a device corresponds to a start ping of the session and a second ping from the device corresponds to an end ping of the session based on (i) the threshold, (ii) first payload data from the first ping, and (iii) second payload data from the second ping. The example apparatus includes the session partitioner to also determine a duration of the session based on a difference between a first time associated with the first ping and a second time associated with the second ping.
1. An apparatus to build a session from ping-level data, the apparatus comprising:
a threshold calculator to determine a threshold based on reference sessions represented in panel data obtained from panel meters; and
a session partitioner to:
determine if a first ping from a device corresponds to a start ping of the session based on (i) the threshold and (ii) first payload data from the first ping;
determine if a second ping from the device corresponds to an end ping of the session based on (i) the threshold and (ii) second payload data from the second ping; and
determine a duration of the session based on a difference between a first time associated with the first ping and a second time associated with the second ping.
2. The apparatus of claim 1 , wherein the threshold calculator determines the threshold based on an average of durations of reference sessions represented in the panel data.
3. The apparatus of claim 1 , wherein the session partitioner is to determine the first ping corresponds to the start ping when at least one of (i) the first ping is an initial ping received from the device, (ii) the first payload data indicates a change in session partitioning, or (iii) a time difference between the first ping and a preceding third ping satisfies the threshold.
4. The apparatus of claim 1 , wherein the session partitioner is to:
determine the second ping is included in the session when a time difference between a third ping and the second ping does not satisfy the threshold, the third ping received between the first ping and the second ping; and
determine the second ping corresponds to the end ping of the session when a time difference between the second ping and a fourth ping satisfies the threshold, the fourth ping received after the second ping.
5. The apparatus of claim 1 , wherein the session partitioner is to:
generate a probability distribution of end durations from panel data;
select a first end duration from a range of possible end durations based on the probability distribution; and
add the first end duration to the second time of the second ping.
6. The apparatus of claim 1 , wherein the session is a first session, and the session partitioner is to determine a plurality of possible sessions from the ping-level data based on different partitioning schemes executed in parallel, the plurality of possible sessions including the first session.
7. The apparatus of claim 6 , wherein a first one of the partitioning schemes is based on a change in genre content in the first payload data, a second one of the partitioning schemes is based on a change in program name in the first payload data, and a third one of the partitioning schemes is based on a change in episode name in the first payload data.
8. The apparatus of claim 7 , further including a session partition selector to select the session from the plurality of possible sessions based on a performance criterion.
9. A method to build a session from ping-level data, the method comprising:
determining, by executing an instruction with a processor, a threshold based on reference sessions represented in panel data obtained from panel meters;
determining, by executing an instruction with a processor, that a first ping from a device corresponds to a start ping of the session based on (i) the threshold and (ii) first payload data from the first ping;
determining, by executing an instruction with a processor, that a second ping from the device corresponds to an end ping of the session based on (i) the threshold and (ii) second payload data from the second ping; and
determining, by executing an instruction with a processor, a duration of the session based on a difference between a first time associated with the first ping and a second time associated with the second ping.
10. The method of claim 9 , wherein the determining of the threshold includes determining the threshold based on an average of durations of reference sessions represented in the panel data.
11. The method of claim 9 , wherein the determining of the duration of the session includes:
generating a probability distribution of end durations from panel data;
selecting a first end duration from a range of possible end durations based on the probability distribution; and
adding the first end duration to the second time of the second ping.
12. The method of claim 9 , wherein the session is a first session, and further including determining a plurality of possible sessions from the ping-level data based on different partitioning schemes executed in parallel, the plurality of possible sessions including the first session.
13. The method of claim 12 , wherein a first one of the partitioning schemes is based on a change in genre content in the first payload data, a second one of the partitioning schemes is based on a change in program name in the first payload data, and a third one of the partitioning schemes is based on a change in episode name in the first payload data.
14. The method of claim 13 , further including selecting the session from the plurality of possible sessions based on a performance criterion.
15. At least one non-transitory computer readable medium comprising instructions that, when executed, cause at least one processor to at least:
determine a threshold based on reference sessions represented in panel data obtained from panel meters;
determine that a first ping from a device corresponds to a start ping of a session based on (i) the threshold and (ii) first payload data from the first ping;
determine that a second ping from the device corresponds to an end ping of the session based on (i) the threshold and (ii) second payload data from the second ping; and
determine a duration of the session based on a difference between a first time associated with the first ping and a second time associated with the second ping.
16. The at least one non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the at least one processor to determine the threshold based on an average of durations of reference sessions represented in the panel data.
17. The at least one non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the at least one processor to:
generate a probability distribution of end durations from panel data;
select a first end duration from a range of possible end durations based on the probability distribution; and
add the first end duration to the second time of the second ping.
18. The at least one non-transitory computer readable medium of claim 15 , wherein the session is a first session, and wherein the instructions, when executed, cause that at least one processor to determine a plurality of possible sessions from ping-level data based on different partitioning schemes executed in parallel, the plurality of possible sessions including the first session.
19. The at least one non-transitory computer readable medium of claim 18 , wherein a first one of the partitioning schemes is based on a change in genre content in the first payload data, a second one of the partitioning schemes is based on a change in program name in the first payload data, and a third one of the partitioning schemes is based on a change in episode name in the first payload data.
20. The at least one non-transitory computer readable medium of claim 19 , wherein the instructions, when executed cause at least one processor to select the session from the plurality of possible sessions based on a performance criterion.