Optimization of broadcast event effectiveness
Introduced herein are methods and systems for determining machine learning marketing strategy. For example, a computer-implemented method according to the disclosed technology includes steps of identifying one or more business metrics to be driven by a marketing plan; generating one or more response functions of the business metrics by performing a machine learning process on a marketing dataset; optimizing a spending subject of the marking plan subject to constraints to generate a marketing strategy based on multiple decision variables; and presenting the marketing strategy to an advertiser.
1. A method performed by a computer system for determining measurable responses to a plurality of broadcast events, the method comprising:
storing, in a memory of the computer system, a set of historical data including measurements of a key performance indicator (KPI), subsets of the set of historical data include metadata comprising at least a timestamp, a channel, and a magnitude to be associated with a given broadcast event of the plurality of broadcast events;
identifying, in a computer system, an existence of a regime switching point for the given broadcast event, the regime switching point based on a cutoff point established by a dynamic threshold difference in performance of the KPI in a recent time period versus a past time period;
wherein:
in response to identification of the existence of the regime switching point:
ignoring, using the computer system, measurement of the KPI before the regime switching point of the given broadcast event, and
crediting, using the computer system, measurement of the KPI in the recent time period, after the regime switching point, to the given broadcast event to predict future performance;
in response to identification of no existing regime switching point, crediting, using the computer system, the measurements of the KPI across the set of historical data to the given broadcast event to predict future performance; and
calculating, in the computer system, a predicted effectiveness of a future broadcast event having a same character as the given broadcast event based on the credited measurements of the KPI used to predict future performance.
2. The method of claim 1 , further comprising:
calculating, by the computer system, a predicted effectiveness of a second future broadcast event having a same character as the given broadcast event based on the measurements of the KPI before the regime switching point and ignoring the measurements of the KPI after the regime switching point.
3. The method of claim 1 , further comprising:
identifying, by the computer system, a beginning cutoff point in time and an ending cutoff point in time for the set of historical data, wherein measurements of the KPI before the beginning cutoff point in time are not credited to the given broadcast event and measurements of the KPI after the ending cutoff point in time are not credited to a second broadcast event.
4. The method of claim 3 , wherein the given broadcast event or the second broadcast event further include a medium, the medium is any of: print; picture; video; audio; or multimedia; and the channel is any of: a website; a television channel; a streaming video or video on demand service; a podcast personality; a radio station; or signage.
5. The method of claim 1 , wherein the dynamic threshold is based on a slope of the measurements of the KPI as a function of time.
6. The method of claim 1 , wherein the given broadcast event further includes an event length and the event length is used as a further basis for said calculating the predicted effectiveness.
7. The method of claim 1 , wherein a budget, a day of week, a national or local availability, an inventory status, an age suitability, or spending timeframe is used as a further basis for said calculating the predicted effectiveness.
8. The method of claim 1 , wherein the regime switching point is identified based on a low point between two peak points in the measurements of the KPI.
9. A non-transitory computer-readable storage medium comprising computer instructions when executed by a processor for determining measurable responses to a plurality of broadcast events, cause a computer system to perform those steps of:
storing, in a memory of the computer system, a set of historical data including measurements of a key performance indicator (KPI), subsets of the set of historical data include metadata comprising at least a timestamp, a channel, and a magnitude to be associated with a given broadcast event of the plurality of broadcast events;
identifying an existence of a regime switching point for the given broadcast event, the regime switching point based on a cutoff point established by a dynamic threshold difference in performance of the KPI in a recent time period versus a past time period;
wherein:
in response to identification of the existence of the regime switching point:
ignoring, using the computer system, measurement of the KPI before the regime switching point of the given broadcast event, and
crediting, using the computer system, measurement of the KPI in the recent time period, after the regime switching point, to the given broadcast event to predict future performance;
in response to identification of no existing regime switching point, crediting, using the computer system, the measurements of the KPI across the set of historical data to the given broadcast event to predict future performance;
calculating a predicted effectiveness of a future broadcast event having a same character as the given broadcast history based on the credited measurements of the KPI used to predict future performance.
10. The non-transitory computer-readable storage medium of claim 9 , further comprising:
calculating a predicted effectiveness of a second future broadcast event having a same character as the given broadcast event based on the measurements of the KPI after the regime switching point and ignoring the measurements of the KPI before the regime switching point.
11. The non-transitory computer-readable storage medium of claim 9 , wherein the dynamic threshold is based on a slope of the measurements of the KPI as a function of time.
12. The non-transitory computer-readable storage medium of claim 9 , wherein the regime switching point is identified as a low point between two peak points in the measurements of the KPI.
13. The non-transitory computer-readable storage medium of claim 9 , wherein the character of the given broadcast event includes an emotional tone and the emotional tone is used as a further basis for said calculating the predicted effectiveness.
14. The non-transitory computer-readable storage medium of claim 9 , wherein the given broadcast event further includes an event length and the event length is used as a further basis for said calculating the predicted effectiveness.
15. A system for assignment of measurable responses to broadcast events, comprising:
a memory of a computer system storing a set of historical data including measurements of a key performance indicator (KPI), subsets of the set of historical data include metadata comprising at least a timestamp, a channel, and a magnitude to be associated with a given broadcast event; and
a processor of the computer system executing instructions stored in the memory that identify an existence of a regime switching point for the given broadcast event, the regime switching point based on a cutoff point established by a dynamic threshold difference in performance of the KPI in a recent time period versus a past time period, wherein existence of the regime switching point causes the processor of the computer system to credit the measurements of the KPI in the recent time period after the regime switching point to the given broadcast event to predict future performance and ignore the measurements of the KPI before the regime switching point, and wherein no existing regime switching point causes the processor of the computer system to credit the measurements of the KPI across the set of historical data to the given broadcast event to predict future performance, and wherein the processor calculates a predicted effectiveness of a future broadcast event having a same character as the given broadcast history based on the credited measurements of the KPI used to predict future performance.
16. The system of claim 15 , wherein the dynamic threshold is based on a slope of the measurements of the KPI as a function of time.
17. The system of claim 15 , wherein the regime switching point is identified as a low point between two peak points in the measurements of the KPI.
18. The system of claim 15 , wherein the character of the given broadcast event includes an emotional tone and the emotional tone is used as a further basis for said determining the predicted effectiveness.
19. The system of claim 15 , wherein the given broadcast event further includes an event length and the event length is used as a further basis for said calculating the predicted effectiveness.
20. The system of claim 15 , wherein the character of the given broadcast event is any of: print; picture; video; audio; or multimedia.