KLM Royal Dutch Airlines: using Machine Learning for optimizing Ad Creatives.
KLM & Squaremoon
KLM found fundamental new insights for improving the CTR and ROAS – and reduced its time spent on finding the right campaign imagery significantly – after they started using the machine learning models behind the Content Analyzer.
The Goal
KLM wanted to find a more data-driven way for finding the right imagery for their campaigns. In addition, they wanted to improve the average performance of their creative assets. For this, they wanted to test if the Content Analyzer could help increase relevancy at scale.
Their Story
KLM Royal Dutch Airlines is the flagship carrier of the Netherlands. Founded in 1919, it is the oldest airline still operating under its original name. In total, KLM offers direct flight services to more than 143 destinations around the globe. To remain relevant and innovative, KLM has always been the frontrunner in introducing new technology to stay connected to the people.
The Solution
KLM collaborated with Facebook Marketing Partner Squaremoon on analyzing the destination imagery in more than 50 countries. Using Squaremoon’s machine learning model, the Dutch airline analyzed all their different creatives and changed the ones with low predicted scores.
Analyzing pictures within the Content Analyzer app.
In the end, it turned out that the visuals with high predicted scores (above 60%) performed more than 69% better than average. Moving on, the explanatory heatmap feature helped KLM to find fundamental new insights for optimizing the content for their advertising channels.
Their success
The Royal Dutch Airline partnered up with Squaremoon to optimize campaign imagery in more than 50 countries. So far, the results look really promising. The ad creatives with high predicted scores have a CTR that is more than 69% higher than average. Also, we found positive relations for other metrics such as CPC and ROAS.
In addition, the partnership helped KLM to significantly reduce time spent on finding the right creative. It now only costs about a couple of minutes to analyze batches with hundreds of images. In addition, the Content Analyzer automatically puts all the images and videos in folders with the help of smart labeling and color detection. This made it possible to find really specific imagery in only a couple of clicks:
The search function makes it possible to find the labeled images and videos in only a couple of clicks.
Quotes
“We already knew that content can have a big impact on ad performance, but given that we mostly work with automated ads and feeds with hundreds of destination images, it was not easy to decide where and how to start. Squaremoon’s tooling has made this content analysis scalable, helping us select which content to replace and even pre-testing our new content. We are excited to see what kind of results this will bring in our campaigns.”
Metten de Vries
“The latest artificial intelligence technologies enabled us to create the Content Analyzer; the smartest image interpreter on the market. Using a scalable setup we make this technology accessible for everyone. The Content Analyzer is especially useful dealing with large image databases, and was found to be helpful for KLM. The tool enabled them to always quickly find the most adequate imagery.”