Cart abandonment rates have been climbing steadily since 2014, after reaching an all-time high in 2013. In 2023, the share of online shopping carts that is being abandoned reached 70 percent for the first time since 2013. This is an increase of more than 10 percentage points compared to the start of the time period considered here. Mobiles vs. desktops When global consumers shop online, they spend considerably more when doing so on desktop computers. In December 2023, the average value of e-commerce purchases made through desktops was approximately 159 U.S. dollars. Purchases completed on mobiles and tablets were of comparable values, ranging between 100 and 105 U.S. dollars. Even though consumers spent more when conducting their shopping on computers, they were more inclined to add products to their shopping carts when using mobile devices. Ultimately, mobile devices provide a convenient and more accessible way to shop, but desktop computers remain the preferred choice for more expensive purchases. Where do consumers shop online? Across the globe, digital marketplaces are shoppers’ number-one online shopping destination. As of April 2024, some 29 percent of consumers voted marketplaces as their favorite e-commerce channel, followed by physical stores and retailer sites. Looking at which retailers’ global shoppers prefer to shop at, amazon.com emerged as the world's most popular online marketplace, based on share of visits. The U.S. portal accounted for around one-fifth of the global online marketplace's traffic in December 2023. Amazon's German and Japanese portal sites ranked third and fifth among the leading online marketplaces, further demonstrating Amazon's dominance over the market.

The OTTO session dataset is a large-scale dataset intended for multi-objective recommendation research. We collected the data from anonymized behavior logs of the OTTO webshop and the app. The mission of this dataset is to serve as a benchmark for session-based recommendations and foster research in the multi-objective and session-based recommender systems area. We also launched a Kaggle competition with the goal to predict clicks, cart additions, and orders based on previous events in a user session.

For additional background, please see the published OTTO Recommender Systems Dataset GitHub.

Key Features

• 12M real-world anonymized user sessions
• 220M events, consiting of clicks, carts and orders
• 1.8M unique articles in the catalogue
• Ready to use data in .jsonl format
• Evaluation metrics for multi-objective optimization

Dataset Statistics

Train/Test Split

Since we want to evaluate a model's performance in the future, as would be the case when we deploy such a system in an actual webshop, we choose a time-based validation split. Our train set consists of observations from 4 weeks, while the test set contains user sessions from the following week. Furthermore, we trimmed train sessions overlapping with the test period, as depicted in the following diagram, to prevent information leakage from the future:

https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4621388%2F94cead9aec2ef687490b1212e40f409a%2Ftrain_test_split.png?generation=1676645044801713&alt=media" alt="Train/Test Split">