Video hosting and sharing services enable creators and advertisers to create campaigns that engage viewers. To price the advertisements, and to give advertisers on the campaign an idea of the popularity of the content, the viewership is predicted. Both under- and over-prediction of views are associated with penalties, respectively of wasted inventory and capacity crunches. View estimations based on channel average suffer from sample bias and invisible trends. This disclosure describes techniques of in-flight view prediction, e.g., predictions of views done after the launch of a campaign for the remaining days of a campaign. The predictions of the total views on a line-up of in-flight videos are based on the distributions of prior view history. The described predictor delivers continuously improving predictions for live videos, and enables determination of whether a campaign is meeting view goals. It thereby enables real-time fine-tuning of inventory and capacity for the remaining days of the campaign.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Mulford, Brian; Ridder, Michael de; Ranger, Colby; Gaffney, T. J.; and Mithani, Jay, "Predicting Content Views Using Finite Integrals", Technical Disclosure Commons, (December 10, 2020)