Abstract: A method may include receiving training data including tuning data, household member data, and initial person level distributions. The method may further include aggregating the tuning data from one or more user devices associated with a household to generate an observed household distribution, and calculating, via a prediction model, an implied household viewership distribution based on the person level distribution associated with one or more members of the household. The method may further include comparing the implied household distribution to the observed household distribution of the household, adjusting the prediction model and/or the person level distributions such that the implied household distribution more closely aligns with the observed household distribution, and generating a report with the person level distributions. A system and a non-transitory computer-readable medium may perform the method.

Abstract: Predict whether active viewability is taking place and/or the likelihood that active viewability will be taking place in the future is described. Historical viewing data may be obtained. One or more probability distribution functions may be generated based on the historical viewing data. One or more survival curves may be determined based on the one or more probability distribution functions. Current viewing data may be obtained. Whether viewability is active in at least one of a current time or a specified future time may be predicted based on the one or more probability distribution functions and the current viewing data. Alternatively, this prediction may be performed via a machine learning model trained on the historical viewing data. Whether to perform a function may be determined based on the prediction.

Abstract: Data indicative of a plurality of content output sessions associated with a first time period may be received. Based on the received content output session data, a first subset of the plurality of content output sessions that are outliers may be determined. Data associated with a device associated with each of the first subset of content output sessions may be received. Based on the received device data, each device may be classified as an outlier device or an inlier device. Based on the classification, a second subset of the first subset of content output sessions that are associated with an outlier device may be determined. A duration of each content output session of the second subset may be compared to a duration threshold associated with a second time period. Based on the comparison, a total duration of the content output sessions of the second subset may be adjusted.

Abstract: Determining attributes for a stream of live content (e.g., live-streaming videos) is described. Initially, a stream of live content is received. One or more attributes are then extracted from a first segment and a second segment of the stream. The first segment and the second segment of the stream may have the same predetermined duration and the second segment may at least partially overlap with the first segment. The one or more attributes extracted from the first segment may be transmitted at a first time, such as at a first break in the stream of content. The one or more attributes extracted from the second segment may be transmitted at a second time, such as at a second break in the stream of content. The attributes from the first and second segments may be transmitted to a content provider and/or a creative provider, such as an advertiser.

Abstract: Techniques for projecting household-level viewing events are described herein. Population data may be accessed including classes of a plurality of demographic attributes for households in a market. Representative household units (RHUs) may be generated, and the RHUs may be assigned a class for each of the demographic attributes and a quota based on the demographic attributes of a plurality of panelist households. Each of the panelist households may be assigned to one of the RHUs based on at least one panelist classes matching the classes for respective demographic attributes of the RHU, and the number of matching panelist households assigned to each of the RHU may be based on the quota. Panelist viewing data representing viewing events associated with the panelist household may be accessed. A report may be generated with the classes of the RHUs and the panelist viewing data of the assigned panelist households.

Abstract: Prediction models for managing viewership data are disclosed. An amount of time users are displayed content is initially obtained. The obtained amounts may be for each content distributor that distributes channels, for each of the channels with respect to which sets of content are displayed, for each of the sets that comprises content displayed during past periods, and for each of the displayed content. A set of features associated with each of the displays is obtained and a target period is selected from among the past periods. A model is used to predict an amount of time users were displayed content during the target period, for each of the sets, channels, and distributors, based on the obtained sets of features associated with the displays during the target period and on the obtained amounts for the displays during the past periods that precede the target period.

Abstract: Methods and systems for determining usage are described. Initially, site-centric data and panel-centric data are accessed and pre-processed. Initial usage measurement data is determined based on the pre-processed site-centric data. One or more adjustment factors are determined based on the pre-processed panel-centric data. The one or more adjustment factors are applied to the initial usage measurement data to generate an adjusted usage measurement data. Reports based on the adjusted usage measurement data are generated.

Abstract: Systems and methods for classifying images (e.g., ads) are described. An image is accessed. Optical character recognition is performed on at least a first portion of the image. Image recognition is performed via a convolutional neural network on at least a second portion of the image. At least one class for the image is automatically identified, via a fully connected neural network, based on one or more predictions, each of the one or more predictions being based on both the optical character recognition and the image recognition. Finally, the at least one class identified for the image is output.

Abstract: A method includes receiving audio data from a microphone of a device associated with a user. The audio data includes audio received by the microphone from a first source and from a second source. The audio data is correlated with a first audio signature and a second audio signature. Then, based on the correlation of the audio data with the first audio signature, first content broadcast from the first source is identified. Second content broadcast from the second source is also identified based on the correlation of the audio data with the second audio signature. It is then determined that the device is closer to the first source than the second source based on audio qualities and/or temporal aspects of the audio data. The first content is associated with the user based on the determination that the device is closer to the first source.

Abstract: The systems and techniques described herein measure advertisement effectiveness of behavior-based outcomes (e.g., site visit, number of pages consumed, searches, online and offline transactions). The system implemented an automated model to measure the impact of exposures and impressions on outcomes using uses panel data, cookie-based data, and combinations thereof. The techniques use test and control approach to calculate effectiveness, where the test group are those exposed to a campaign and a control group who is not exposed. For those exposed, a running analysis of impressions (and other variables) in a pre period is used to determine behavior based outcomes over a set time period after that exposure. As a result, the automated model is able to generate metrics that show absolute and relative impacts on future behavior.

Abstract: Audience engagement is analyzed by receiving at least one analysis parameter comprising a content selection parameter, identifying one or more content based on the content selection parameter, determining viewership for the one or more content, determining one or more keywords for the one or more content based on the content, filtering social media messages based on the determined one or more keywords, and calculating an audience engagement measurement corresponding to the one or more content based on the viewership and the social media messages.

Abstract: Systems and methods herein log traffic to and from a device on a network. Logging can occur using a metering device, router, proxy, or other elements. For example, a metering device operatively coupled to a routing device can log the traffic directed to and originating from a user device. Logged traffic can be analyzed to identify users, devices, and/or sessions. For example, an identifier unique to the user device in the session, a device type of the user device, and a specific user of the device during the session can be identified.

Abstract: Webpage or other resource accesses by client systems may be recorded, and those accesses may be analyzed to develop audience measurement reports. At times, it may be desirable to segment those reports according to classes of client systems (e.g., work vs. home client systems). A given client system can be classed into one of the reporting classes based on one or more classes of network service providers that provide the client with access to a network. The recorded resource accesses and classes of the client systems can then be used to generate audience measurement reports that are segmented according to one or more of the client system classes.

Abstract: Data indicative of times at which at least one item of supplemental video content was output via a plurality of client devices may be received. A first time at which the at least one item of supplemental video content was output via a greatest quantity of client devices of the plurality of client devices may be determined. A second time at which the at least one item of supplemental video content was output via a second greatest quantity of client devices of the plurality of client devices may also be determined. A schedule associated with output of the at least one item of supplemental video content may be generated based at least on the first time and the second time.

Abstract: Techniques for projecting household-level viewing events are described herein. Population data may be accessed including classes of a plurality of demographic attributes for households in a market. Representative household units (RHUs) may be generated, and the RHUs may be assigned a class for each of the demographic attributes and a quota based on the demographic attributes of a plurality of panelist households. Each of the panelist households may be assigned to one of the RHUs based on at least one panelist classes matching the classes for respective demographic attributes of the RHU, and the number of matching panelist households assigned to each of the RHU may be based on the quota. Panelist viewing data representing viewing events associated with the panelist household may be accessed. A report may be generated with the classes of the RHUs and the panelist viewing data of the assigned panelist households.

Abstract: Mobile usage data representing the access of one or more resources on a network by mobile devices is accessed. The mobile usage data includes information received from mobile devices as a result of beacon instructions included with the one or more resources. A first set of data representing information about accesses to the one or more resources by mobile devices with persistent beacon cookies is determined. A second set of data representing information about accesses to the one or more resources by mobile devices with non-persistent beacon cookies is determined. One or more adjustment factors are determined based on the first set of data. A count of unique visitors accessing the one or more resources from the mobile devices is determined based on a count of accesses by the mobile devices with persistent beacon cookies and a count of accesses by the mobile devices with non-persistent beacon cookies adjusted by the one or more adjustment factors.

Abstract: Systems and methods herein log traffic to and from a device on a network. Logging can occur using a metering device, router, proxy, or other elements. For example, a metering device operatively coupled to a routing device can log the traffic directed to and originating from a user device. Logged traffic can be analyzed to identify users, devices, and/or sessions. For example, an identifier unique to the user device in the session, a device type of the user device, and a specific user of the device during the session can be identified.

Abstract: Methods and systems are provided herein for calibrating subject data based on reference data, so that the calibrated subject data more closely represents a target population. The methods and systems include partitioning a reference data set into a plurality of reference data partitions using a data partitioning scheme, each reference data partition associated with a characteristic; and partitioning a subject data set into a plurality of subject data partitions using the data partitioning scheme, each subject data partition associated with a characteristic that corresponds to the characteristic associated with a reference data partition of the plurality of reference data partitions; identifying a variable present in the reference data set that is not present in the subject data set; and calculating a value of the variable for each reference data partition based on a rate of occurrence of the variable in each reference data partition.

Abstract: A method of classifying a product placement in a video using rule sets is disclosed. Each rule of the rule set includes a value and one or more defining rule elements. An attribute rule set is created with attribute values and attribute elements that define levels of audio visual prominence of a product in the video. An integration rule set is created with integration values and integration elements where the integration elements define levels of integration of the product with video continuity. The video is partitioned at product scene changes to create product blocks. For each product block, an attribute value is selected based on the attribute elements and an integration value is selected based on the integration elements. An impact parameter for the video is derived as a function of the selected attribute values and integration value.

Abstract: Measuring a networked audience is described. Initially, a first set of network usage data based on access of a resource by a first set of client systems is received. Next, a second set of network usage data based on access of the resource by a second set of client systems using a monitoring application installed on the second set of client systems is determined. Usage of the resource based on the first set of network usage data during a time period and one or more adjustment factors based on the second set of network usage data are determined. The determined usage is adjusted using the one or more adjustment factors. Finally, one or more audience reports for the resource using the adjusted usage are generated.