top of page
  • Writer's pictureDavid Hornstein

There is accounting for taste!

It’s very important for me to understand my clients’ preferences in order to design a home they will find appealing. Of course, every designer faces the same challenge of learning their clients’ taste. How they go about it varies widely. Some designers like to visit a client’s existing home to see the choices they have previously made. I find this only works in cases where the client’s previous furnishings were chosen somewhat recently and fairly deliberately. People’s taste does evolve over time, and many clients purchase “starter” furniture that does not necessarily reflect their current preferences.

Of course, I can simply ask my clients what they like, but I find that it’s quite difficult to describe aesthetic preferences in words. Learning someone’s taste is a challenging problem that I’ve been studying for years.

It's hard to know what these words actually mean

Back in 2000, I started a company (StylePath) designed to understand people’s aesthetic preferences and direct them to products like home furnishings that matched their taste. To do so, I embarked on a multi-year research project to understand if individuals’ aesthetic preferences can be learned and predicted.

I learned that some individuals express their preference in terms of styles like “contemporary”, or “country”, or “traditional”. Others say they are “eclectic” and their taste does not conform to a particular style. However, in both cases, it turns out that what people really like are certain attributes combined in particular ways, that may or may not conform to a defined “style”.

The surprising fact is that no matter how eclectic a person’s taste may appear, it is actually fairly constant, and therefore predictable. In other words, even if you like a combination of attributes that spans across a variety of styles, you will consistently like that same set of attributes.

StylePath used AI to learn user preferences by showing test images and asking the user to rate them. This system attained 80% accuracy- eight out of ten times it could predict a person’s preference for a new item based on their PSP (personal style profile). In a similar vein, many designers ask clients to assemble scrapbooks (either physical or virtual) of things they like. This can be a useful method (a picture is worth a thousand words), but suffers from two limitations:

1) The client may not be aware of the full range of possibilities

2) What they like in a particular image may be misunderstood. I learned this early in my career when I used to show a slideshow of my previous work to clients and ask them what they liked and disliked. I was often surprised to find that people liked very different styles, until I learned that what they really were responding to was the light coming in the window in each image, and not the actual design!

The patented StylePath process uses sets of images that are carefully chosen to test for certain aesthetic attributes and not others. The entire image set represents the full range of aesthetic possibility. I ask clients to respond to images that they either love or hate. This process is then repeated with a refined image set. While this takes a bit of time up front, most people enjoy the process. More importantly it saves time and money over the course of the entire project, as the hundreds (or thousands) of decisions that go into the design and furnishing of a home are all informed by the client’s PSP, which greatly simplifies the selection process.

Recent Posts

See All


bottom of page