View-Based Ad Placement in Scrollable Units
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional online systems inaccurately credit impressions for advertisements in scrollable advertisement units, leading to decreased revenue due to user alienation and varying advertisement interaction probabilities, as all ads in a unit are credited regardless of user interaction, even if users do not navigate through them.
Innovation Solution
An online system determines advertisement positions within a scrollable advertisement unit based on predicted performance and bid amounts, using discount factors and expected values to rank and place ads, ensuring that higher-value ads are displayed in more visible positions, and charges advertisers accordingly based on interaction probabilities and positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple advertisements are combined into a scrollable advertisement unit to reduce display area, then user alienation is reduced, but the system cannot accurately track which advertisements are actually viewed
Solution Approach 1:
The system pre-calculates discount factors for each position in the scrollable advertisement unit based on historical interaction data. These discount factors represent the probability that an advertisement at a given position will be viewed by the user. By performing this calculation in advance, the system can accurately attribute impressions even though the exact viewing behavior is unknown at display time.
Solution Approach 2:
The system uses historical user interaction data with scrollable advertisement units to continuously refine the discount factors for each position. When users actually scroll through and interact with advertisements, this feedback is used to update the predicted performance metrics, improving the accuracy of impression tracking over time.
2Productivity
If all advertisements in a scrollable unit are credited with impressions regardless of user interaction, then revenue is maximized, but advertisers are charged inaccuracyly for ads not viewed
Solution Approach 1:
Instead of applying a uniform crediting rule to all advertisements in the scrollable unit, the system applies position-specific discount factors to each advertisement. Advertisements at positions with higher predicted visibility receive higher crediting weights, while those at less visible positions receive lower weights. This local differentiation allows accurate reflection of actual viewing probabilities.
Solution Approach 2:
The system dynamically adjusts the impression crediting parameter (discount factor) based on the advertisement's position and predicted performance. Rather than using a fixed crediting rule, the parameter changes according to position-specific metrics derived from historical data, enabling both accurate tracking and optimized revenue.
3Productivity
If advertisements with high bid amounts are always placed in visible positions, then revenue is maximized, but the system cannot account for advertisement-specific properties that affect user interaction
Solution Approach 1:
The system pre-calculates expected values for each advertisement by combining the bid amount with position-specific discount factors and advertisement-specific properties. This preliminary calculation creates a comprehensive metric that accounts for all factors before placement decisions are made, ensuring both high revenue and accurate property consideration.
Solution Approach 2:
The placement decision uses a composite metric (expected value) that combines multiple factors: bid amount, position discount factor, and advertisement-specific properties. This composite approach integrates diverse elements into a single ranking criterion, allowing the system to simultaneously optimize for revenue while accounting for various ad characteristics.
Data Source
AI summary
An online system selects advertisements for inclusion in a scrollable advertisement unit that includes a display area and multiple advertisements, each associated with a position in the scrollable advertisement unit. Positions in the scrollable advertisement unit are ranked based on a measure of predicted performance of an advertisement in each position. Advertisements are ranked based on a probability of being viewed by a user if associated with a particular position in the scrollable advertisement unit and, optionally, on a probability of presenting an advertisement based on characteristics of the advertisement. The position ranking and the advertisement ranking are used to associate advertisements with positions. For example, an advertisement is associated with a position having a location in the position ranking corresponding to the advertisement's position in the advertisement ranking.