De kom, de klickade, de övergav

Sammanfattning: This thesis analyses web statistics belonging to a national B2C e-commerce in retail. The purpose of the thesis is to investigate abandoned carts amongst logged in customers with at least one product added in to their cart and at least one visit to the payment page. Since the response variable "cart" is binary and we are studying the probability of a cart being abandoned, the suitable method of choice is the logistic regression model. The findings conclude that the majority of the customers are women between 30-60. A linear expected probability was found when investigating a correlation between age and abandoned cart. Those under 30 years of age had a 70% probability of abandoning the cart whereas those over 70 years of age had a 30% probability of leaving the cart. Only a small portion of the customers were new to the company. Which proved to be of more significance to the outcome of the cart, than if the customer had been a frequently recurring customer in the past. That is - without including the customers with extreme values in Orderhistory ("Orderhistorik") where a exceedlingly high conversion rate could be seen. After considering age, the only channel proven to have a significant influence on the customer cart, was the "SMS" channel. Both the amount of time spent on the site and the amount of pages scrolled through had a significant positive effect on the probability of a cart being abandoned. The day that the customer entered the web page also had a significant impact. "Day 5" had the highest probability of abandoned carts. However, it seemed like the variation correlated with different behaviors in the age groups. The most frequently added products were from the categories "Ladies" (Dam) and "Home interior" (Heminredning). However, products from the category "Dam" were expected to be less abandoned than "Heminredning". The number of products also had a significant impact on the cart, but since most customers added between 1 and 3 items to their carts, the relevance of the finding is perhaps low. When investigating the "Checkout" page, where customers process the purchase, a few findings were made. The number of times a customer visited the "Checkout" page was significant to the outcome of the cart. When leaving the page, it seemed like the customer most frequently went back to look for more products - since the product pages for "Dam" and "Rea" correlated with "Checkout". When considering the size of the cart, the probability increased. However, there were no interactions between the two of them. The payment method was found to be of moderate importance, where the least used method also was the most abandoned one. An error message related to a discount was shown to 20% of the customers. When this message occurred the cart had a 20% probability of being abandoned.