Learn how the online marketplace for used cars processes more than 5.5 million ads daily with Keboola
Tomas Trnka, Chief Data Officer at EAG group (Carvago's parent company), shares how Carvago turned a simple idea into a company with $7M annual revenue in just three months.
Every day, Carvago analyzes 5.5 million car ads in destinations all over Europe, offering a smooth car shopping experience across borders for everyone.
Carvago, a digital marketplace for buying and selling used cars, offers the best car deals to over 15,000 B2C and B2B customers without owning a single vehicle. The secret? Data.
Running Carvago without a data team is impossible. Carvago sells data as its product and is among the top 3 AWS clients in the Czech Republic. We process 5.5 million ads and 600,000 new photos daily. This includes different types of data processing from image recognition to natural language processing and other machine learning models to classify, filter, and enrich the ads.
Additionally, Tomas revealed that manually processing a car ad takes about an hour. To handle millions of car ads at this volume, Carvago would require a staggering 500,000 employees.
Carvago realized the classic approach would delay their launch by years.
Keboola was the weapon of choice from the very beginning. One day we need 1 hour of computing, then the next 1,000 hours. We needed a solution that scales up and down. With Keboola, we can focus on business growth without DevOps concerns the ads.
We evaluated other options, yet none matched Keboola's comprehensive capabilities. Keboola is an all-in-one platform. It gives you out-of-the-box monitoring, orchestrations, multiple backends, storage, and data governance.
When Carvago was building its data product, Keboola took care of the DevOps, data warehouse, scaling resources, and even setting up all of the administrative backend. This allowed the team to focus on building the Carvago platform, rather than worrying about the DataOps.
Keboola's agility enabled us to transform a business idea into a production-ready MVP within three months, a process that might have taken years with a traditional data management approach.
By automating machine-learning tasks, Keboola enables Carvago to process 5.5 million ads every day. The scale of Carvago's operations reveals the critical role of automation. However, automation extends far beyond ad processing.
Tomas and his team of nine data experts built a system to automate numerous business operations, including standardizing bank receipts for control, scoring prospects in Salesforce, and automating procurements for logistics.
A person can manually process 1 car ad per hour to make it ready for our website. So 8 ads per day. We’re processing 5.5M ads daily at Carvago. Without Keboola, we would need 560,000 people to do the same job.
Keboola centralizes all data assets and jobs in a single platform, under a unified data governance framework. This enables Tomas and his team to quickly reuse previous work for launching new data use cases.
This efficiency not only speeds up development, but also ensures consistency and quality across projects.
Here’s an example of the platform's impact:
For a new marketing project, I can immediately reuse our current work—accessing relevant data sets and transformations, and identifying the last contributor with ease.
Keboola simplifies data integration with features like 700+ pre-built connectors and a drag-and-drop visual flow builder. This ensures quick data synchronization across a diverse range of sources.
Tomas shares the effect of Keboola on their workflows:
With Keboola, we can get data from various places in minutes, transform it, and send it elsewhere. You can automate the whole process in an instant.
Keboola’s rapid integration capabilities underscore its role as an indispensable asset for dynamic data management.
Tomas and his team cooperate with the Czech Technical University on electrical car grant projects.
In one key project, they're working with students to harness the power of machine learning and refine a price predictor for cars, drawing on vast historical ad data to do this - a step towards more accurate market insights.
Traditional data stacks often falter at sharing extensive historical data; they can be hampered by volume constraints and elevated risks of security breaches, especially with external partner access.
Using Keboola, Tomas and his team effortlessly launched a new project and securely shared the extensive historical data with university students, this time without the usual concerns over governance and security breaches.
Thanks to Keboola’s versatility
... we were able to develop our entire Data Product within its ecosystem. This approach eliminated the need for us to design an appropriate stack from various tools, allowing us to concentrate on enhancing the quality of our output and reducing time to market.
Thanks to Keboola
... we dared to try and launch Carvago as a business, which was a crazy idea at the time.