Data Science & PPC Case Study. Machine Learning and Paid Search

Forward3D was tasked with solving challenges associated with large scale paid search for an ecommerce client.






The Problem



PPC requires advertisers to make quick investment decisions based on historical performance activity across a large portfolio of keywords linked to products. Marketers want to funnel investment into their best performing products in order to deliver the best returns. 



But how do you do this effectively if you have tens of millions of unique products on offer at any one time, each with very little performance history or brand recognition? 



And how is this challenge for intelligent optimisation complicated further on a peer-to-peer ecommerce site where stock of specific products can deplete rapidly, new products are continually introduced and the product set is so wide that search term selection becomes difficult.



Our Objective



Forward3D was tasked with creating a PPC optimisation strategy to help an international ecommerce client drive more revenue within their ROAS targets, whilst overcoming the challenges described above.

With the client in an aggressive growth phase and striving to achieve tough targets across multiple markets, successful revenue growth was imperative.

Key Achievements


After six weeks of implementing our solution, the client had achieved the following:


  • 22% increase in keyword volumes resulting in significantly increased product coverage
  • A 6.1% increase in total revenue achieved at an ROI above the client's target
  • Achievement of the client’s revenue targets in one market just one week after the strategy introduction
  • Another key success was that they were able to confidently invest in high return DSA (Dynamic search ad) campaigns without fear of risk to their brand due to the anomaly detection system.


How Did We Do It?


To read how Forward3D solved the problems that the client was faced with through a machine learning-based solution, please complete the form below.