Abstract: The present invention uses a combination of image decomposition, behavioral data and a probability engine to provide products, which are closest to a consumer's personal preference. This personal preference is referred to as “taste-based technology”. Taste-based technology uses three key components: image analyzer, behavior tracking and predication engine. The image analyzer uses a number of techniques to decompose an image into a number of image signatures, and then places those signatures into a database for later analysis and retrieval. The techniques used include: storing geometric descriptions of objects of the domain, which are matched with extracted features from the images; processing data from lower abstraction levels (images) to higher levels (objects); and processing data that are guided by expectations from the domain. This decomposed data can be used as standalone data (i.e. in a non-web environment) or fed into the prediction engine for real-time consumer preference determination.