Schibsted Marketplaces: Powering smarter recommendations with Bigtable and BigQuery
Reduced infrastructure costs by 70% compared to previous solution
Improved data access speed and enabled faster model development, improving time to market
Increased potential for future AI integration
Schibsted Marketplaces, a leading online classifieds group in the Nordic region, cut infrastructure costs by 70% and accelerated data insights and model development by adopting Bigtable and BigQuery. This led to faster, more relevant recommendations and a better user experience.
Schibsted Marketplaces, a leading online classifieds group in the Nordic region, cut infrastructure costs by 70% and accelerated data insights and model development by adopting Bigtable and BigQuery. This led to faster, more relevant recommendations and a better user experience.
Enhancing efficiency and reducing costs with Bigtable
Online marketplaces now account for the largest share of online purchases worldwide. While this is due in part to a pandemic era ecommerce boom, the continued growth of marketplaces over the past three decades is rooted in strong buyer-seller communities that increase with each new member. Recommendation engines are central to this dynamic, helping buyers discover relevant products while boosting seller visibility.
"Recommendations are a crucial aspect of online marketplaces. They drive traffic and help buyers find what they’re looking for—or even discover something they didn’t know they wanted—therefore enabling vendors to sell their products faster and more easily," says Johan Boberg, Engineering Manager at Schibsted Marketplaces. "They’re an essential part of keeping users engaged."
Schibsted Marketplaces, which is currently in the process of shifting its workloads to Google Cloud, operates in Norway, Sweden, Finland, and Denmark, focusing on four core verticals: mobility, real estate, employment, and "re-commerce," or second-hand goods. The company previously based its recommendation engines on a distributed, non-relational database but found themselves between tiers in its scaling model, leading to inefficiencies and high costs with a non-cloud-native hosting setup. The team decided to shift to Bigtable, enabling them to significantly reduce overall costs, improve performance and boost operational efficiency.
"Bigtable turned out to be a great alternative in terms of cost and also allowed us to keep our services in one place, simplifying our operations by reducing the need to manage multiple vendors," says Boberg.
Bigtable allowed us to keep our services in one place, simplifying our operations by reducing the need to manage multiple vendors. It offers greater real-time consistency, ensuring that deletions propagate across all instances immediately, making it more reliable. But the major difference has been in terms of cost. Since the shift, we’ve saved around 70%.
Johan Boberg
Engineering Manager, Schibsted Marketplaces
Powering fast and easy data analysis with BigQuery
Recommendation engines depend upon machine learning (ML) models that predict what content or products a user might be interested in by analyzing their behavior and patterns. For the past 10 years, data scientists at Schibsted worked with data stored in files, which complexified and slowed down model development. In order to streamline data access and reduce retrieval times, the team implemented BigQuery.
"Moving to BigQuery has significantly improved the speed and ease with which our data scientists can do their job," says Boberg, "This allows them to develop models faster, experiment more, and ultimately deliver better products to our users sooner. We now have a much shorter time to market."
The team worked closely with Google Cloud experts based in Norway to implement and optimize the new solutions. "We've had a continuous dialogue with the Google team, and whenever necessary, they've called in product experts to help us design our setup and ensure we’re using it in the most efficient way," Boberg adds. "Instead of just letting us transfer over from our previous solution, they reviewed our specific use case and made sure we were on the right track. I think that was a key factor in our success."
Moving to BigQuery has significantly improved the speed and ease with which our data scientists can do their job. This allows them to develop models faster, experiment more, and ultimately deliver better products to our users sooner. We now have a much shorter time to market.
Johan Boberg
Engineering Manager, Schibsted Marketplaces
Looking to the future with Vertex AI
Boberg is excited about the possibilities offered by Vertex AI when it comes to building, training, and using ML models. "Right now, our data scientists are using Vertex AI to develop some of the models we use, but it’s very much a work in progress," he explains. "We’re also using it for other machine learning models beyond recommendations. It's helping with faster model development, and I think there's a lot of potential there."
Boberg and his team are enthusiastic about the potential of AI and ML to transform not only their recommendation engine, but also the development of entirely new products and possibilities within Schibsted’s marketplaces. "The pace of development in AI is extremely exciting. It feels like a paradigm shift. We’re seeing the emergence of completely new products and possibilities based on this technology that we couldn’t have imagined before," he says. "As far as we’re concerned, it’s crucial to stay on top of these developments. It’s not just about top-down directives; everyone across the teams needs to see how they can take advantage of AI."
While it’s still early days, the Schibsted team is looking forward to continuing to collaborate with Google Cloud throughout this process. "Having our services under the Google Cloud umbrella also opens up potential to create more innovative products for our customers in the future," Boberg says.
Schibsted is a family of digital brands with more than 5 000 employees with world-class media houses in Scandinavia, and leading marketplaces and digital services that empower consumers. Millions of people interact with Schibsted companies every day.
Industry: Retail
Location: Norway
