When we speak of recommendations systems in a business context, we tend to speak about them in terms of their direct, measurable financial impact. This is an important consideration, of course. It would make little sense to invest time, energy and money in a complex set of tools if the product of that hard work amounted to nothing more than window dressing. At the same time, when we adopt this mindset, we fail to recognize some of the benefits that accrue to the folks who see the products of our fancy mathematical models. Recommendations systems can make people happier by sorting through the mass of products available, or find what they’re looking for more quickly.
On face, online commerce’s capacity to offer an expansive catalog — far broader than anything a physical store could — is an enormous advantage. But it can also undermine the store: the huge number of choices can overwhelm people. Faced with an array of 2,000 different televisions or 4,000 different pairs of shoes, people who set out to update their home theater or wardrobe often give up when they cannot make heads or tails of the huge number of choices. This phenomenon, the so-called “paradox of choice” or “choice overload” entered the zeitgeist through The Paradox of Choice, due to Swarthmore College professor Barry Schwartz. It has been demonstrated in numerous contexts. Schwartz himself co-authored a study that supported his theory in the context of Web search, and other research has documented choice overload in contexts as wide ranging as the choice of gourmet chocolates to the choice of a retirement plan.
This choice paralysis obviously has direct, measurable business implications for online retailers: people buy fewer products, or they don’t buy anything at all. But it’s not as if shoppers leave this situation happy not having spent the money. That may be the case in some circumstances, of course; someone whose shopping trip was short-circuited by choice overload will be happy if it prevented him from buying a giant television he couldn’t afford. More often than not, however, we’re disappointed when we abandon our hunt for a new shirt or mobile phone in frustration, an intuition supported by research. Most of our shopping has a purpose, and when we “fail” in this respect, we’re frustrated.
Recommendations, personalized to a given shopper’s preferences, offer a solution. By quickly identifying the much smaller set of products a given shopper might want to begin with, one no longer faces the often paralyzing, frustrating task of sorting through dozens, hundreds or even thousands of different choices. It’s the difference between choosing among a few similarly-styled frocks one is likely to buy — a choice between, say, purple or puce — and wading through a vast pool of choices, most of which are irrelevant. Urban fashionistas no longer have to hunt for avant-garde designer wares buried among conservative basics, and trial attorneys no longer have to skip past pair after pair of pre-distressed jeans in their hunt for a well-made pair of wool trousers.
When done well, personalized recommendations help retailers and shoppers alike. Retailers get a gentle mechanism to suggest the products shoppers are likely to find interesting. At the same time, shoppers have an easier time navigating the bewildering variety of products that modern technology allows online retailers to offer. It’s a win-win.
Infinite Analytics has always been at the forefront of cutting edge predictive analytics and semantic technologies. Our Natural Language Processing (NLP), Machine Learning and Semantic Technologies help us establish relationships between users, brands, stars who endorse those brands, movies, affiliation to social causes & prediction of voter inclination.