Skip to main content Skip to footer
Case study

The challenge

A digital advertising agency struggled to improve click-through and conversion rates for the ad extensions it created for its customers. Its existing manual process, as well as the efforts to keep pace with customers’ dynamic web content and the constant monitoring added to the challenge and hampered speed to market. The agency engaged Cognizant to improve its campaign conversion rates.

Our approach

The manual process the agency initially followed included writing compelling call out text and maintaining word counts, character limits and the number of extensions added. Adhering to the inconsistent ad extension platform guidelines resulted in a slow process. 

Cognizant's media & entertainment automation experts designed an automation tool to create ad extensions that feature deep learning-based natural language processing (NLP). Based on Python, this tool helps determine the best ad placements for each customer’s website, improving productivity and effectiveness. The solution includes five key components:

  • Collecting real-time insights on the company’s brand offerings through website scraping
  • Clustering and converting relevant text into ad extensions
  • Processing the text through a recurrent neural network with a long short-term memory deep learning model to predict the selection probability of new words
  • Recommending terms with the highest probability of selection
  • Tuning the hyper-parameters to ensure model accuracy 

Based on artificial intelligence (AI), the new solution has helped enhance the performance of ad extensions across a large number of websites and improved the productivity of campaign managers by 50%.

Implementing AI and advanced analytics increases ad extension effectiveness

Cognizant’s AI-based solutions and integration of advanced analytics into the agency’s existing manual process has helped the company not only improve campaign managers’ productivity but also increase ad extension effectiveness by 10%. The solution’s scalable and reusable methodologies and algorithms can be used in multiple NLP scenarios. 


productivity gains


accuracy of deep learning algorithm


increase in click-through rates