Physical Ad Effectiveness Estimation
Both the data platform and automatic report generation product got fully deployed in production.
Both the data platform and automatic report generation product got fully deployed in production.
Goal
Create robust data platform to support multiple machine learning use cases. Estimate and predict ad impact and automatically generally comprehensive reports about physical ad campaign effectiveness.
Impact
Both the data platform and automatic report generation product got fully deployed in production.
In the first year, campaign effectiveness estimation generated 400K in ARR
Automated report generation from: ~ 2 weeks to 1 day
Inference time pe 1 billion rows: ~ 3 minutes
Problems
Estimating the impact of and the number of views for any billboard ad campaign using GIS data of the sampled Link’s panel.
Solution
A data platform and a data lake with functioning ETL and data aggregation from all sources.
Machine learning model to predict whether an outdoor ad was seen and data processing to estimate the general effectiveness of any ad campaign on given Swiss demographics.
Technical Highlights
2-step solution that optimizes the way the schedule is created based on the problem size that resulted in >90% reduction in runtime compared to the benchmark and no loss in schedule quality.
Testimonial
“The benefit of working with Prophecy Labs is based on their close collaboration between staff members, a very communicative approach and open mindset, a diverse team and a willingness to perform in any situation. I highly recommend Prophecy Labs for data science and data engineering projects, especially for bringing successfully Proof Of Concepts to the next stage.”
Sylvestre Lucia, Product Manager at Link Institute
Discover even more recent use cases where complex challenges were met with innovative solutions, resulting in tangible business value.