Back in September, Amazon launched Scout, its tool for recommending furniture and home decor items. It works by prompting users for “thumbs up” or “thumbs down” ratings of selected items and then, Amazon says, by using machine learning algorithms to make recommendations based on thousands of visual attributes.
Then, two weeks ago, we found that Amazon had integrated a Scout-like recommendations module on furniture product pages that works in a similar way.
Now, we are seeing a third product variant of the Scout tool, called the Home Style Quiz. It’s a 6-step quiz that results in furniture and home decor product recommendations, and is now live on Amazon.com in the US. Here’s how it looks:
Step 1. Amazon gives users 15 pictures of fully decorated rooms, and asks, “Which three looks are your favorites?”
Steps 2, 3, and 4. Amazon chooses items based on the user’s input in the previous step, and prompts users to “Vote on at least three pieces to help us narrow down your unique style.” A total of 20 items are presented in each of these three steps.
Step 5. Amazon asks users to express preferences on colors. Amazon asks, “Let’s talk color–how much is just right?” Users can choose between three options: “Keep it neutral,” “Some pops of color,” or “The bolder the better.”
Step 6. Finally, users are asked, “Lastly, how much pattern and texture are perfect?” Available answers are: “None, thanks,” “Less is more,” and “Can’t get enough.”
The ensuing results page offers suggested products grouped into three categories. In our example, the results were: “Your unique blend of styles is Industrial, California Casual, and Glam.” A section for each of those categories is listed below, offering recommended products in each section. Here’s what the “Industrial” section looks like in our example.
Finally, there are “Color & Pattern Recommendations” and “Products You Liked” sections.
Big picture, Amazon is continuing to experiment with new ways of exposing users to machine learning services to improve its personalization and recommendation algorithms. The Scout concept makes sense for highly visual product categories like furniture and home decor. We believe more intelligent personalization is an area in which Amazon can improve significantly through further R&D.
Amazon has chosen not to invest heavily in human recommendations much, a la how StitchFix manually curates for apparel. However, Amazon has created a little-known tool called Outfit Compare, in which customers can upload photos of themselves and get an Amazon staffer’s opinion of which looks better.