Lucky Cart: The power of data analysis for consumers, retailers and brands

Thanks to Google Cloud, Lucky Cart can now processes very large volumes of data and performs the complex calculations required to set up ultra-personalized marketing operations for its customers.

Lucky Cart analyzes receipt data collected in real time to predict shopper consumption and optimize brand presences for their partners. They use the data to gain a better understanding of shoppers’ individualized purchasing profiles, to enable an ultra-personalized e-retail and increase profits.

Results of Google Cloud implementation and Devoteam G Cloud support:

• Supports the delivery of ultra-personalized customer activations, by enabling Lucky Cart to handle large volumes of data
  
• Empowers the development team by keeping operations to a minimum with autoscaling on Google Kubernetes Engine
  
• Helps support growth by delivering a reliable infrastructure, so the tech team can focus on adding new services

Retailers are always searching for new ways to increase consumer engagement online, and one way to do this successfully is through gamification. In one study, introducing an element of competition or challenge helped content and media companies increase online customer interactions by 30-40%, and the potential applications of this approach are endless.

This is one of French marketing technology company Lucky Cart’s key strategies and it’s transforming the way fast-moving consumer goods (FMCG) brands and distributors are leveraging data to increase customer engagement. With access from between 10 to 20 million lines of purchase receipts a day through its network of brands and distributors, Lucky Cart uses artificial intelligence to create ultra-personalized marketing campaigns, in which shoppers can try their luck at a promotional game to win back the total value of their shopping cart. It works with 85% of the FMCG market in France, and is the only player in ecommerce media with this level of distribution.

“Shoppers see an integrated banner on the distributor’s platform encouraging them to purchase multiple goods from a certain brand, and if they fulfill the purchasing criteria, they are invited to play the game,” explains Nicolas Trannoy, Marketing Director at Lucky Cart. “As well as broad and static targeting, our solution enables ultra-personalized and dynamic profiling, so each shopper is treated as an individual. For the distributors, that means a 20% increase in value of the average basket.”

To deliver this level of ultra-personalization, Lucky Cart handles large volumes of complex first-party data, creating profiles using sophisticated algorithms and then handling browsing data in close to real time. This data is then fed back into the system, to refine and improve future predictions. With the goal of improving its capacity for handling very large volumes of data, Lucky Cart decided to migrate its infrastructure to Google Cloud.

“What enables us to stay ahead of our competitors is the work of our data science team,” says Trannoy. “The added value we offer is the capacity to generate ultra-personalized profiles, and that’s why we chose Google Cloud for our data infrastructure.”

Choosing a cloud provider to support big data analyses:

To deliver a high level of personalization based on constantly refining its knowledge of consumer habits, Lucky Cart analyzes its distributors’ data to create shopper profiles, always in compliance with GDPR regulations. As this includes three years of historic sales data, the volumes involved are very large, and its previous system was not able to provide sufficient computational power to carry out the necessary requests.

“Our previous system was not sufficiently scalable, in terms of the machine learning and deep learning capabilities we require,” explains Julien Guitard, Chief Data and Science Officer at Lucky Cart. “The infrastructure was not very intuitive to use, and the lack of scalability was holding back both our data engineers, and our data scientists.”

To enable it to access the power it required to carry out complex calculations for every shopper, across hundreds of categories, two years ago Lucky Cart switched both its application services infrastructure, and its data analysis infrastructure to Google Cloud.

Keeping the focus on development, not operations

Its Kubernetes workloads are now orchestrated with Google Kubernetes Engine to run all its application services, with related files, including for the Lucky Cart web page, stored in Google Cloud Storage. It also uses Google BigQuery as part of its infrastructure data pipeline.

“Google Kubernetes Engine is easy to configure, and scales really well,” says Vincent Oliveira, CTO at Lucky Cart. “That means the developers don’t need to think about managing it in production, they can simply set the parameters and be confident it will work.”

Lucky Cart also uses managed services, such as Google Cloud Functions, and plans to continue leveraging more Google Cloud tools for automation, having already tested Google Cloud Run for one-off projects to make app development and deployment faster and easier. “As Google Cloud tools are easy to configure, we’re able to have only one or two people focusing on Operations within a technical team of around 20,” Oliveira adds.

Crunching millions of transactions daily

The data science team at Lucky Cart, led by Guitard, is responsible for managing the Lucky Cart data analysis, including data structuring, extract, transform, and load (ETL) processes, and automating models and queries developed by data scientists to create new tools, products and services.

The team uses BigQuery to store and manage the multiple data sets, as well as Dataproc to manage the computations to calculate personalized discounts. It also uses a number of tools, including Cloud Functions, Pub/Sub and Cloud Build to connect and communicate between services, and Google Marketing Platform to create reports.

Lucky Cart leverages BigQuery for its incoming and outgoing data, storing the service history from its customers and what has been displayed for each shopper in one data set. Another data set stores all the computations based on this data, such as discounts and A/B testing results, and additional data sets are created for prospective partners and new customer samplings.

“With around three million customer transactions a day, BigQuery helps us to handle very large sets of data, including around 3TB of customer transaction data, and up to 70TB of computations based on that data,” says Guitard. “Over two years of using it, we’ve never had any issues with even our biggest queries.”

Continuing to grow with Google Cloud, Devoteam G Cloud and MongoDB

Lucky Cart participated in the Google for Startups Cloud program and MongoDB for Startups to help the company accelerate its business and scale their data needs. MongoDB for Startups is a selective program for early-stage, high-growth startups. Designed to help founders and CTOs build and grow data-driven organizations, its program provides a wide range of resources, including MongoDB’s data platform, personalized technical advice, and access to its robust developer community.

Lucky Cart employed MongoDB Atlas to help its data scientists with their detailed work. MongoDB Atlas is a multi-cloud application data platform. It has an integrated suite of ​​cloud database and data services to accelerate and simplify how businesses build with data.

With Google Cloud and MongoDB Atlas, the data science team at Lucky Cart can now carry out complex computations that previously would not have been possible. “By enabling us to scale our data infrastructure, machine learning, and artificial intelligence. Google Cloud and MongoDB are supporting our ability to deliver a fully personalized experience to the end-customer, which is key to keeping our brands and distributors happy,” explains Guitard.

Lucky Cart also worked side-by-side with Google Cloud Partner, Devoteam G Cloud, to accelerate Lucky Cart’s migration. Devoteam G Cloud provided Google Cloud expertise during the initial project phases and is providing Google Cloud training to technical teams at Lucky Cart to support onboarding.

Google Cloud has also supported the business in scaling up, as Lucky Cart has grown the number of distributors it works with from one to 10 over the past two years. “Our landscape has changed significantly, and the fact that we’re able to look forward to adding more distributors without worrying about whether the solution will continue to work is a big enabler for us,” says Trannoy.

The data science team is now testing out Cloud Composer to access more dynamic workflows, and is also in the process of putting Vertex AI into production to complement its current Dataproc approach for its machine and deep learning models. “This will enable us to work at scale, and to manage additional complexity as we expand,” says Guitard.

On the business side, Lucky Cart is now looking to grow beyond its French customer base, and is also planning to create new products to personalize the entire customer journey, beyond its promotional games offer. “We know that for any new project, we’ll be able to find a tool on Google Cloud that answers our needs,” says Oliveira. “That means we’re confident that we’ll be able to easily scale up our infrastructure from ten to hundreds of distributors over the coming years.”