Online used car retailer CarNext was confronted with a unique challenge in relation to Google Ads: Google did not have enough conversion data to work with, which meant that the ads on Google could not be optimized in the most efficient way. As a result, advertising costs were high and results were disappointing. Cmotions was called in for advice and assistance. Together, the companies worked on developing a fully automated data science solution so that Google can better optimize the ads and performance. With the deployment of the new solution, CarNext can now advertise more efficiently and achieve more engagement.
CarNext is an innovative company that sells used cars online. The pioneering nature of its business model makes it difficult to use Google Ads. Google uses the available conversion data to continuously optimize the results of advertisements. But if there is not enough data, this optimization mechanism of Google does not work.
“For the optimization of our ads and finding the right audience, Google did not have enough conversion data,” says Senem Kirac, Insights & Analytics Manager at CarNext. “After all, online car buying is still in its infancy. As a result, we spent large amounts on Google Ads without getting the desired results. Due to the lack of data, there was no artificial intelligence behind the display of the ads, so the cost per lead was above average. As Google could not help us, we decided to find a solution ourselves.”
More data was needed to optimize the ads and to find the right audience. “We examined which activities, such as making an appointment or asking a question, correlated with a transaction,” Senem explains. “We attached a certain value to each activity. We shared this information with Google so that the ads and the audience could be optimized. This led to an improvement, but there was still not enough data.”
CarNext called on the knowledge and expertise of Cmotions to significantly increase the number of data points. “Together with Cmotions, we started to analyse the behaviour of visitors who came to our website via Google Ads,” says Senem. “We did this with a Markov Chain model, a common technique used to analyse attribution. Based on historical data from Google Analytics, Cmotions created and trained the data science model. This enabled us to learn which activities contribute to people making a (showroom) appointment or buying a car. Previously, we did this on the basis of gut feeling. Now we do it on the basis of data.”
The results were massive. Whereas previously CarNext collected only a few hundred data points per month, this now shot up to 600,000. This is exactly the data Google needs to optimize the ads. “Based on the model, a feedback loop was created for Google,” Senem explains. “All visitors who come to the website via Google Ads show certain behaviour. The trained model maps out this behaviour and assigns scores to it. We send this scores back to Google via Search Ads 360. This provides Google with feedback on the traffic they send to our website and optimizes our ads.”
During the project, the CarNext team worked closely with the Cmotions team. There was also good contact with Google. “For Google, this was also a new problem, so they were keen to help find a solution,” says Senem. “Cmotions was the driving force behind the project and steered the cooperation between different parties, such as the performance marketers, data engineers and other colleagues within CarNext. Cmotions was also responsible for building, testing, optimizing and automating the model. This was no easy process.”
“There was a moment when the performance was disappointing,” Senem continues. “As we could not provide hard evidence that the model worked, it was difficult to convince senior management of its added value. Cmotions handled this very well. We received regular and extensive feedback on the results. Thanks to clear presentations, we knew exactly what the status of the project was and what we could expect. Cmotions thus ensured that the project did not get shut down when it was hanging by a thread, but that everyone kept believing in it. That is quite a challenge during such an innovative project as this.”
In order to be able to say something about the results of the model, the return on ad spend (ROAS) was examined. “Each time we wanted a higher return, the system was effortlessly able to follow. The result is largely dependent on the input of the performance marketers, but it is extremely satisfying that this system enables us literally to be able to turn the knobs. Now we need to further optimize performance using the model. We are now only using the model in France. If we are completely satisfied with the results, we will also roll it out in other countries.”
A major added value of the project is that the model not only enables CarNext to spend the marketing budget for Google Ads more efficiently, but also to better understand customer behaviour.
"The model gives us an insight into which activities on the website contribute to making an appointment or buying a used car. All visitor behaviour can be accurately mapped and this enables us to better understand customer behaviour. We can adjust our marketing activities accordingly and thus better respond to the needs of potential customers. The added value of this model has already been demonstrated."