Work

Mikä suosittelutyyppi tuottaa todennäköisimmin ostotapahtuman, kun sitä on jo klikattu?

Year: 2016
Type: dissertation
Abstract: Wide variety of surveys have studied recommendation systems (RS) and the topic is very current. However, the general research topics are related on what algorithm is the best or how algorithms differ from each other and what happens if different algorithms are combined. Also customers’ attitudes and trust (including privacy issues) toward RS are surveyed. This study investigates what happens after a certain (history-related, item-related or toplist) recommendation type is clicked, which will produce the most carted items and the most importantly purchased items. The data used was from a company providing recommendation services for online stores, thus the company sells only the recommendation service itself. The data was from a real life web store providing very valuable and rare knowledge. Because of the company’s clients’ privacy policy, the source of data was not known. Known is that the web store was large European store and most customers were European too. An important remark is that the data was gathered only from one web store and included one million single sessions. The data was studied and linear regression models were contributed. Also binary mode format was used to exclude certain occasions. Results gave some strong indications that there is a notable difference between recommendation types. Re-sults were also connected with a literature findings and psychological persuasion, informative per-suasion and clickstream behavior were discussed. The history type was clearly the most effective recommendation type. Item related recommendation was the second most effective. The reason of the difference is the better and more precise knowledge of the user for the history related recommendation. Toplist was clearly the weakest type to provide purchased items. It is totally impersonal, so the informative aspect is very low, only providing knowledge of popular items. Findings suggest that everything is based on to the knowledge of the user. More accurate recommendations will provide more purchases. As men-tioned the amount of recommendation researches is very large, yet certain areas are unexpectedly unstudied. More attention should be assigned to practical side of the recommendations and as this study suggests it can provide a great benefit.
Author: Mäkinen, Sami
Institution: Aalto University
Language: English
Open Access status: green
OpenAlex ID: W7112419393
Cites: 73
Cited by:
Related to: 10