Ad Forecasting Workflow Using Historical Index Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing forecasting methods and systems for web-based advertising campaigns suffer from inaccuracy, leading to inefficient energy consumption and carbon footprint, as they fail to provide precise predictions for ad impressions and clicks.
Innovation Solution
A computer-implemented method and system that utilizes a forecasting service to translate targeting criteria and placement data into an index querying syntax, query historical searches, and derive click-through-rate (CTR) in real-time, using AWS Lambda and Apache Druid for fast data processing, and monitor resource consumption to improve forecasting accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional forecasting methods are used for web-based advertising campaigns, then the system can provide forecasts, but the forecasting accuracy is poor leading to inefficient energy consumption
Solution Approach 1:
The system pre-calculates and stores forecasted impressions and clicks in advance using historical data and machine learning models. This preliminary forecasting allows the system to provide accurate predictions without performing complex real-time calculations during ad delivery, thereby reducing computational energy consumption while maintaining high forecasting accuracy
Solution Approach 2:
The system creates simplified copies or approximations of complex forecasting models that can be executed quickly. By pre-processing historical data and storing aggregated statistics, the system can provide accurate forecasts with minimal real-time computation, reducing energy usage while maintaining measurement precision
2Speed
If traditional forecasting methods are used, then the system can operate, but the response time is slow preventing real-time forecasting
Solution Approach 1:
The system performs data aggregation, model training, and forecast calculation in advance before real-time queries are made. Historical data is pre-processed and stored in optimized formats, allowing the system to return forecasts in real-time (within seconds) without sacrificing accuracy, as the heavy computational work was completed preliminarily
Solution Approach 2:
The forecasting system is divided into separate components: data collection, historical analysis, model training, and real-time query processing. Each component operates independently and can be optimized separately, allowing the real-time interface to respond quickly while the background processes maintain forecasting accuracy through thorough analysis
3Reliability
If inaccurate forecasting is used, then the system requires more impressions to compensate, but this increases energy usage and bandwidth requirements
Solution Approach 1:
The system implements feedback loops where actual ad performance data is continuously collected and used to refine forecasting models. This feedback mechanism improves forecast reliability over time, allowing the system to predict impressions and clicks more accurately and reduce the number of impressions needed to achieve campaign objectives, thereby reducing energy waste
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is disclosed a computer-implemented method of serving an advertisements campaign, the method including the steps of: (i) a platform which programmatically authors, schedules and monitors workflows, receiving data by extraction and transformation from an Events database, the platform including an advertisements inventory, and an advertisements Forecasting Inventory directed acyclic graph (DAG); (ii) evaluating and storing advertising indexes in a data & report engine, the advertising indexes including data ingested from the advertisements Forecasting Inventory DAG in the platform; (iii) a data streaming platform receiving pushed forecast events from an advertisements forecasting service; (iv) populating the Events database from the data streaming platform; (v) a portal requesting an advertisements forecast from the advertisements forecasting service; (vi) the advertisements forecasting service querying inventory, by querying the advertising indexes stored in the data & report engine; (vii) the forecasting service using the response to querying inventory to provide an advertisements forecast to the portal; (viii) the portal processing the received advertisements forecast to determine an advertisements campaign; (ix) the portal writing the determined advertisements campaign to a datalayer; (x) the datalayer dumping the determined advertisements campaign to a web service interface; and (xi) the web service interface using the dumped determined advertisements campaign to get campaign content from the platform, including from the advertisements inventory, to serve an advertisements campaign. Related methods and systems are disclosed.