What makes us trust a recommender system?
E-commerce companies could benefit from considering insights from qualitative studies when developing their recommender systems. In a study at the University of Gothenburg, the writers aim at giving companies the tools to better understand their customer’s perspective.
Recommender systems are an important part of today’s e-commerce. Filtering out products from large online catalogues could be an overwhelming task and makes the recommendation function a crucial key.
Maria Saxborn and Yuechen Pan, Masters in Communication alumni and the writers of the article Trust Through Recommendation in E-commerce, have based their study on the recommendation system used by Zalando by interviewing visitors of their website. The ambition is to help companies who value their customers’ shopping experience to improve their recommender systems. To do that, they decided to extend the so called “Trust Building Model”, used by developers of e-commerce and recommender systems to understand what makes a customer feel trust in the service offered.
Trust makes us stay
“The higher trust, I would say, the more beneficial it is for the e-commerce.”, says Maria Saxborn.
“As a consumer, if you feel like it’s a trustworthy site, that could give good recommendations and it’s connected to what you have looked at before, that will most likely lead you to stay on the site, keep searching for items and maybe even purchasing something.”
Authenticity and quality
The writers found that customers perceive the service as more authentic when they have immediate access to customers’ reviews and can see photos of influencers using the clothes, rather than standard studio photos of models. The feeling of authenticity is one of the new aspects that they have been able to point out as having an effect on trust building towards the recommendation.
The recommendation quality, that is, how well the suggestions match your interests and taste, is another parameter for trust building.
“You will receive a few recommendations when visiting the website, but the first few are the most important, because if they are not accurate, your trust will decrease. If the recommendation matches your interests and your style, you will have more trust. So it’s very important for companies to see that part – how you get as accurate as possible on the first few recommendations”, says Yuechen Pan.
A new kind of study
Historically, recommender system research has focused on how to make the systems more effective, measured by for instance page traffic and purchases, but the consumers’ experiences of online shopping with the support of these systems and their feelings towards them, has not been examined to a large extent before.
“Previous research has not necessarily entailed interviews and qualitative studies on the consumers. This is important to make the models we have today more applicable for the recommender systems”, says Maria Saxborn.
Text: Agnes Ekstrand
Recommender systems are algorithms designed to suggest products or services to users based on their preferences and behavior. In e-commerce, these systems analyse for instance past purchases and browsing history, to offer personalised recommendations.
