Completion Time Estimation

Summary

The client is a bookseller also having online bookstore, ebooks, magazines, music, movies, toys & games. The company popular and have physical retail outlets in the United States and a retailer of content, digital media, and educational products

Understanding the Challenge

The Customer has a dedicated IT infrastructure and team that handles all the digital marketing. The marketing campaigns are diverse and are comprised of automated programs that ran over extensive durations of time on data from Google BigQuery. These programs process different amounts of data and are interdependent. The marketing campaigns have weekly recurring deadlines and that’s why it is highly significant to estimate the accurate completion time of the campaigns.

Solution

By utilizing the Artificial Intelligence powered solution with-in the Machine Learning realm, our solution uses partial data from the past executions to estimate the completion-time based on the number of records to be processed, number of API calls to be made and some other categorical variables and utilizes Random Forest Regression techniques to reach an accurate estimate.

• Accuracy: For a batch of campaigns that span over 20 hours, the estimation is within ±42 minutes’ error margin. This is very much helpful in assessing the value and effectiveness of the marketing campaigns.
• Continuous Improvement: The algorithm improves over the time through Machine Learning. As with preceding time the more campaigns are executed, hence more data is collected. Which in turn helps the algorithm to improve its performance accordingly.
• Automation: Monitoring the campaigns was a manual hectic work before this solution. The automation relieves the pain of manual monitoring and sends email notifications in case of any breakdown or malfunction.
• Campaign Scheduling: The availability of an accurate estimate for the completion time helps in scheduling new campaigns with low error margin and more precision. This helps the utilization of the available time to the full potential. Which eventually helps in better engagement rate.

Result

Thanks to the efforts of Folio3, the solution has enabled customer’s Digital Marketing Team to substantially improve the effective delivery through precise schedules and outcome of marketing campaigns, while meeting the weekly associated deadlines.

Technology used: Scikit Learn, NumPy, Pandas