Sell more longtail with image search

Many common forms of personalization lead retailers straight into the bestseller bubble. Based on a rather narrow data base, many shops suggest products to their customers that many people have bought anyway.

The result is a loop that is as ineffective as it is infinite: Customers buy bestseller products because they are frequently suggested. Bestseller products are frequently suggested because they are bought a lot. Thus, numerous items beyond the bestseller hit list fall by the wayside - the so-called longtail.

To break out of this bestseller bubble in the traditional way, retailers would have to invest a lot of effort and data.

A quick calculation: Reliable predictions for purchase recommendations require at least 100 purchases and more than a thousand views per product. Thus, a relatively small portfolio of 5,000 products would need five million views in a few days, evenly distributed across all product pages, to operate effectively. Very few shops have access to such extensive and representative data.  

Visual Search allows online retailers to break out of this bubble easily and efficiently. The hit rate for recommendations increases significantly, ensuring that individual items do not become shelf warmers.

Our customer Siemens Schuhcenter confirms this positive effect on longtail. Items are "displayed at the right time, increasing their chance to be purchased".

‍