Pinfluencer Adds Image Recognition Tech to Pinterest Analytics Platform

Firm eventually plans to add SKU-level matching

Pinterest attribution can be complicated for marketers. The main issue is that someone could pin a product image from one page on a brand’s site, but that page could feature multiple product images. A typical measurement dashboard would only be able to link the pin and the site page, but Pinterest analytics firms like Curalate and now Pinfluencer have begun to roll out image-level analytics.

Pinfluencer announced today an image-recognition engine that would let it be able to tie a pinned image with an image on a brand’s site and comes in addition to Pinfluencer’s URL matching. Pinfluencer CEO Sharad Verma explained how it works.

“The way our crawlers work is we pull data from Pinterest continuously, extracting pins, pinners, comments, likes. Then we start separating pins that [belong] to a brand’s website. That’s what we’ve been doing until now. Then we start separating images for the webpages. At the end of the whole crawl process and data processing and image processing, here are the images that got pinned, the products that got pinned and here’s the website performance,” he said.

That attribution means a retailer like Sephora (a Pinfluencer client) could see which lipstick in particular is driving Pinterest traffic to its "lips" category page. The beauty giant could then take that insight and promote that lipstick on its home page, create its own pinboard featuring the product and complementary items or add in-store signage identifying it as popular on Pinterest.

The idea is similar to competitor Curalate’s technology, which was launched last spring, but Verma pointed out Pinfluencer’s URL matching as a point of differentiation. “If you don’t do URL analytics the way we do it, what Curalate ends up doing is they don’t know if an image lives on your website or another’s brand’s website,” he said, though it’s Adweek’s understanding that Curalate is in fact able to access the URL for which page a product image resides on.

Pinfluencer’s image-recogntion technology doesn’t have to remain limited to Pinterest. The company has hired a new chief architect from Google who previously helped to build Yahoo’s document search engine, Verma said, adding that Pinfluencer’s technology is similar to search engine technology. “We can go to any site like Google’s crawlers and parse text and images,” he said, specifying eBay, Zappos and Walmart as three potential examples of sites it could crawl in a similar way to Pinterest. “A particuar use case where we might do image recognition beyond Pinterest is for a [consumer packaged-goods] brands. For example, Tide might want to know how its products are doing on eBay, Pinterest, Fancy [etc.]. That’s where we would use image recognition to identify their products,” he said.

Multiple times Verma emphasized that image recognition is only one technology that’s an important piece of a larger puzzle also includes URL matching. Eventually Pinfluencer plans to add SKU-level matching, meaning the ability to match a product’s size and color with a pin.