Adaptive Cost Estimation for Marketing Campaigns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cost estimation methods for online marketing campaigns are volatile for low event rates and sluggish for high event rates, failing to maintain optimal responsiveness and robustness simultaneously.
Innovation Solution
Adaptive cost estimation is achieved by relating the estimator gain to the temporal volatility of the cost estimate, allowing for dynamic adjustment based on estimated revenue rate and events per revenue, with constraints on minimum and maximum responsiveness to balance volatility and response time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed estimator gain is used for cost estimation, then the estimation is simple to implement, but the cost estimate becomes volatile for low event rates and sluggish for high event rates
Solution Approach 1:
The patent applies dynamics by making the estimator gain adaptive rather than fixed. The gain is dynamically adjusted based on the estimated revenue rate and events per revenue, allowing the system to automatically adapt to changing event rates. This resolves the contradiction by enabling the estimator to maintain appropriate responsiveness across different event rate conditions without increasing implementation complexity significantly.
Solution Approach 2:
The patent changes the parameter of estimator gain from a fixed value to a dynamically adjusted value based on system conditions (revenue rate and events per revenue). This parameter change allows the system to optimize its performance across different operating conditions, preventing both volatility at low event rates and sluggishness at high event rates while maintaining reasonable implementation complexity.
2Speed
If the estimator gain is increased to improve responsiveness, then the response time decreases, but the cost estimate becomes more volatile
Solution Approach 1:
The system dynamically adjusts the estimator gain based on the estimated revenue rate and events per revenue. When event rates are low, the gain is reduced to prevent volatility; when event rates are high, the gain is increased to improve responsiveness. This dynamic adjustment resolves the contradiction by allowing the system to have high responsiveness when needed while maintaining stability when event rates are low.
Solution Approach 2:
The estimator gain parameter is changed from a static value to a dynamically adjusted parameter that responds to system conditions. This allows the system to optimize the trade-off between responsiveness and volatility by adjusting the gain based on actual performance data, achieving both fast response when appropriate and stability when needed.
3Stability of the object's composition
If the estimator gain is decreased to reduce volatility, then the cost estimate becomes more stable, but the response time increases
Solution Approach 1:
The estimator gain is dynamically adjusted based on the estimated revenue rate and events per revenue rather than being fixed at a low value. This allows the system to maintain stability when event rates are low (by using a lower gain) while quickly responding when event rates are high (by increasing the gain), thus avoiding the permanent loss of responsiveness that would result from a consistently low gain setting.
Solution Approach 2:
The estimator gain parameter is adjusted based on system conditions to optimize the balance between stability and response time. When conditions warrant faster response, the gain is increased; when stability is more important, the gain is decreased. This conditional parameter adjustment resolves the contradiction by allowing the system to have both stability and responsiveness at different times as needed.
Data Source
AI summary
Embodiments of the present invention provide systems, methods, and computer storage media directed at cost estimation. In embodiments, a method may include receiving an observed event volume and an observed revenue in the current logical interval for a campaign. The method may also include tracking state information of various revenue state variables and events per revenue state variables associated with the previous logical interval. In addition, the method may include determining a cost estimate for a present logical interval of the campaign based on the state information, the observed event volume, the observed revenue, and an events per revenue forgetting factor. Other embodiments may be described and/or claimed herein.