Estimating Unique Ad Impressions via Probability Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating unique ad impressions in online video distribution systems are often inaccurate and inefficient, leading to uncertainties in ad inventory management, resulting in unanticipated surpluses or shortages of ad inventory.
Innovation Solution
A method involving the determination of a discrete probability distribution of video advertising segments per unit time, random sampling without replacement, and the use of a binary search algorithm to estimate the number of unique client devices exposed to ads, allowing for precise forecasting of unique ad impressions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If prior methods are used for estimating ad inventory, then the estimation process is simple, but the accuracy and reliability of the estimate is poor
Solution Approach 1:
The system performs preliminary actions by collecting and storing ad delivery data from multiple client devices before the actual estimation is needed. This pre-collected data forms the basis for generating accurate probability distributions, allowing the system to provide reliable estimates when queries are made without having to conduct new measurements at estimation time.
Solution Approach 2:
The patent introduces probability distributions as an intermediary mathematical model between the raw ad delivery data and the final unique impression estimates. This intermediary layer allows the system to handle the complexity of converting total ad deliveries into unique viewer estimates while maintaining accuracy and enabling efficient querying.
2Measurement precision
If sophisticated estimation methods are implemented, then the accuracy of unique ad impression estimates improves, but the computational efficiency and processing time deteriorates
Solution Approach 1:
The system pre-computes probability distributions from historical ad delivery data and stores them for rapid retrieval. When estimation queries are received, the system efficiently generates unique impression estimates by sampling from these pre-established distributions, avoiding the need for complex real-time calculations and achieving both accuracy and speed.
3Measurement precision
If random sampling is performed multiple times to improve estimate accuracy, then the precision of the estimate increases, but the time and computational resources required increases
Solution Approach 1:
The system performs a sufficient number of random samples to achieve the desired level of estimation accuracy, but not excessive samples that would waste computational resources. The probability distribution framework allows the system to determine the appropriate sampling depth needed to confidently estimate unique impressions from the pre-collected ad delivery data.
Data Source
AI summary
Estimating a number of unique ad impressions in a streaming video system includes defining parameters of an ad campaign and a desired number of ad impressions for the campaign. A computer system determines a discrete probability distribution of video advertising segments per unit time per client device in a population of video advertising segments streamed to a plurality of client devices, based on historical data. The system randomly samples the probability distribution without replacement, based on the defined number of desired ad impressions. An enhanced binary search algorithm may be used for the sampling. Each sample of the probability distribution identifies a number of ads streamed to a different client device in the probability distribution. The system determines, based on the sampling, a number of unique client devices included the samples, thus obtaining an estimate of unique ad impressions for the defined ad campaign.