Self-consistent inception architecture for efficient baselining media creatives

Abstract

A self-consistent inception architecture includes a process that integrates online data and offline data to determine an estimated lift (“prior lift”) in the number of unique visitors (UVs) to a website caused by a television (TV) spot airing on an offline medium. The prior lift is used to adjust a UV profile of the website. A baseline thus produced is fitted through an inception process in which a locally weighted scatterplot smoothing algorithm is applied iteratively until a final baseline converges. The baseline from the inception process is used to determine a calculated lift. If the prior lift and the calculated lift are not consistent (e.g., within a threshold), the process is run iteratively until the prior lift and the calculated lift are consistent. The calculated lifts can be used to determine and visualize performance metric(s) relating to media creatives such as TV spots airing in the physical world.

What is claimed is: 
     
         1 . A self-consistent baselining method, comprising:
 receiving, by a computer, online data from one or more spot airing data providers, the online data relating to a television (TV) spot airing on an offline medium, the online data comprising user activity on an entity’s website and associated with a particular time when the activity occurred;   receiving, by the computer, offline data from the one or more spot airing data providers, the offline data relating to a television (TV) spot airing on an offline medium, wherein the offline data is associated with the entity and is associated with a particular time that the TV spot aired;   determining, by the computer based on the online data and offline data, an estimated lift in a number of unique visitors (UVs) to a website caused by the TV spot airing on the offline medium;   adjusting, by the computer, a temporal response pattern observable from website traffic data, wherein the temporal response pattern relates to a number of UVs at the entity’s website on a minute-by-minute basis within a time window after the TV spot airing on the offline medium;   fitting, by the computer, a baseline through an inception process in which a locally weighted regression algorithm is applied iteratively until a final baseline converges;   computing, by the computer utilizing the final baseline, a calculated lift in the number of UVs to the website caused by the TV spot airing on the offline medium;   comparing, by the computer, a previously calculated lift and the calculated lift;   iteratively performing, by the computer, the adjusting, the fitting, the computing, and the comparing until the previously calculated lift and the calculated lift are consistent;   determining, by a performance analyzer utilizing the calculated lift consistent with the previously calculated lift, one or more performance metrics relating to the TV spot airing on the offline medium;   generating, by a visualizer, visualizations based on the one or more performance metrics; and   presenting to a user, by the computer through a user interface, a dashboard displaying the generated visualizations based on the one or more performance metrics.   
     
     
         2 . The self-consistent baselining method of  claim 1 , wherein the online data comprises the number of unique visitors to the website over a period of time during which the TV spot airing on the offline medium, wherein the offline data comprises spot airing time and spend associated with the TV spot, and wherein the estimated lift is determined utilizing at least the spend associated with the TV spot. 
     
     
         3 . The self-consistent baselining method of  claim 1 , wherein the adjusting comprises subtracting the previously calculated lift from a UV line to thereby generate the baseline for the inception process, the baseline including the temporal response pattern. 
     
     
         4 . The self-consistent baselining method of  claim 1 , wherein the locally weighted regression algorithm is a locally weighted scatterplot smoothing (LOWESS) algorithm. 
     
     
         5 . The self-consistent baselining method of  claim 1 , wherein the inception process comprises applying the algorithm continuously until a final baseline converges to a constant line. 
     
     
         6 . The self-consistent baselining method of  claim 1 , wherein the comparing further comprises, responsive to a determination that the calculated lift is not within a threshold of the previously calculated lift, setting the calculated lift as the previously calculated lift. 
     
     
         7 . The self-consistent baselining method of  claim 1 , wherein the calculated lift is computed by subtracting the final baseline from the number of UVs to the website. 
     
     
         8 . A system, comprising:
 a processor;   a non-transitory computer-readable medium; and   stored instructions translatable by the processor for: 
 receiving online data from one or more spot airing data providers, the online data relating to a television (TV) spot airing on an offline medium, the online data comprising user activity on an entity’s website and associated with a particular time when the activity occurred; 
 receiving offline data from the one or more spot airing data providers, the offline data relating to a television (TV) spot airing on an offline medium, wherein the offline data is associated with the entity and is associated with a particular time that the TV spot aired; 
 determining, based on the online data and offline data, an estimated lift in a number of unique visitors (UVs) to a website caused by the TV spot airing on the offline medium; 
 adjusting a temporal response pattern observable from website traffic data, wherein the temporal response pattern relates to a number of UVs at the entity’s website on a minute-by-minute basis within a time window after the TV spot airing on the offline medium; 
 fitting a baseline through an inception process in which a locally weighted regression algorithm is applied iteratively until a final baseline converges; 
 computing, utilizing the final baseline, a calculated lift in the number of UVs to the website caused by the TV spot airing on the offline medium; 
 comparing a previously calculated lift and the calculated lift; 
 iteratively performing the adjusting, the fitting, the computing, and the comparing until the previously calculated lift and the calculated lift are consistent; 
 determining, utilizing the calculated lift consistent with the previously calculated lift, one or more performance metrics relating to the TV spot airing on the offline medium; 
 generating visualizations based on the one or more performance metrics; and 
 presenting to a user, through a user interface, a dashboard displaying the generated visualizations based on the one or more performance metrics. 
   
     
     
         9 . The system of  claim 8 , wherein the online data comprises the number of unique visitors to the website over a period of time during which the TV spot airing on the offline medium, wherein the offline data comprises spot airing time and spend associated with the TV spot, and wherein the estimated lift is determined utilizing at least the spend associated with the TV spot. 
     
     
         10 . The system of  claim 8 , wherein the adjusting comprises subtracting the previously calculated from a UV line to thereby generate the baseline for the inception process, the baseline including the temporal response pattern. 
     
     
         11 . The system of  claim 8 , wherein the locally weighted regression algorithm is a locally weighted scatterplot smoothing (LOWESS) algorithm. 
     
     
         12 . The system of  claim 8 , wherein the inception process comprises applying the algorithm continuously until a final baseline converges to a constant line. 
     
     
         13 . The system of  claim 8 , wherein the comparing further comprises, responsive to a determination that the calculated lift is not within a threshold of the previously calculated, setting the calculated lift as the previously calculated. 
     
     
         14 . The system of  claim 8 , wherein the calculated lift is computed by subtracting the final baseline from the number of UVs to the website. 
     
     
         15 . A computer program product comprising a non-transitory computer-readable medium storing instructions translatable by a processor for:
 receiving online data from one or more spot airing data providers, the online data relating to a television (TV) spot airing on an offline medium, the online data comprising user activity on an entity’s website and associated with a particular time when the activity occurred;   receiving offline data from the one or more spot airing data providers, the offline data relating to a television (TV) spot airing on an offline medium, wherein the offline data is associated with the entity and is associated with a particular time that the TV spot aired;   determining, based on the online data and offline data, an estimated lift in a number of unique visitors (UVs) to a website caused by the TV spot airing on the offline medium;   adjusting a temporal response pattern observable from website traffic data, wherein the temporal response pattern relates to a number of UVs at the entity’s website on a minute-by-minute basis within a time window after the TV spot airing on the offline medium;   fitting a baseline through an inception process in which a locally weighted regression algorithm is applied iteratively until a final baseline converges;   computing, utilizing the final baseline, a calculated lift in the number of UVs to the website caused by the TV spot airing on the offline medium;   comparing a previously calculated lift and the calculated lift;   iteratively performing the adjusting, the fitting, the computing, and the comparing until the previously calculated lift and the calculated lift are consistent;   determining, utilizing the calculated lift consistent with the previously calculated lift, one or more performance metrics relating to the TV spot airing on the offline medium;   generating visualizations based on the one or more performance metrics; and   presenting to a user, through a user interface, a dashboard displaying the generated visualizations based on the one or more performance metrics.   
     
     
         16 . The computer program product of  claim 15 , wherein the online data comprises the number of unique visitors to the website over a period of time during which the TV spot airing on the offline medium, wherein the offline data comprises spot airing time and spend associated with the TV spot, and wherein the estimated lift is determined utilizing at least the spend associated with the TV spot. 
     
     
         17 . The computer program product of  claim 15 , wherein the adjusting comprises subtracting the previously calculated from a UV line to thereby generate the baseline for the inception process, the baseline including the temporal response pattern. 
     
     
         18 . The computer program product of  claim 15 , wherein the locally weighted regression algorithm is a locally weighted scatterplot smoothing (LOWESS) algorithm. 
     
     
         19 . The computer program product of  claim 15 , wherein the comparing further comprises, responsive to a determination that the calculated lift is not within a threshold of the previously calculated, setting the calculated lift as the previously calculated. 
     
     
         20 . The computer program product of  claim 15 , wherein the calculated lift is computed by subtracting the final baseline from the number of UVs to the website.