It is a data coop that pools anonymized data about online visitors’ purchase intents, which is then used by advertising platforms for ads or other targeting. Participating site publishers share the data coop’s revenue. Skimlinks says that its network covers 1.5 million websites, about 20,000 retailers and 1.4 billion shoppers.
When it was announced in November of 2015, Skimlinks said that the visitor data was divided into segments of 200 or so product categories, with indications about how soon the users in a given segment wanted to buy that product.
As it turns out, CEO and co-founder Alicia Navarro told me, those “product categories” were really about product-specific retailers and brands, not actual products.
Let’s say a user browses to a page on a participating publisher’s site. The publisher might have a review of new Sony cameras. Since Sony makes other kinds of electronic products, though, the “product category” might be consumer electronics. If the page was on, say, the site of a retailer that specializes in cameras, then the category might be cameras.
Navarro told me that her company’s machine learning was generating some segments in categories that were actual products, like microwave ovens, but there were too few categories, with too little user data, to be useful.
This week, however, Skimlinks is announcing it can now generate actual product categories — such as “digital cameras” — at scale, creating about 900 product categories. They include electric irons, golf bags, humidifiers, microwave ovens and homeopathy.