Montreal’s Propulse Analytics is a new startup that arms retailers with deep learning technology to improve product recommendations for shoppers. The company raised $1.8 million in seed funding (US $1.4 million).

The funding comes from Wobemail Online Services LLP and Stradigi Ventures. Frank & Oak was one of Propulse’s early customers.

Propulse’s technology analyzes the desires and intent of consumers and provides accurate product choices for them. The technology also factors in purchase and browsing history in order to provide “the most timely and accurate product choice available.”

The startup is “leveraging the features of product images to better grasp customers’ tastes and ultimately increase sales for retailers.”

The company said it starts by understanding consumers’ tastes where others simply look at purchase and browsing history. Its engine analyzes product images while considering 39,000 variables. It looks into historical data on browsing interest and purchase patterns, offers the best-matching products in real-time and looks at inventory to determine what’s available, or what will be available.

Techcrunch’s John Mannes revealed that Propulse CEO Eric Brassard used to work at Saks Fifth Avenue, where savvy sales people regularly earned $200,000 on the shop floor.

Within humans, the art of selling clothing is recognizing the tastes and personalities of customers quickly and recommending appropriate items, noted Mannes.

Ideally we can mimic this aspect of clever human psychology within artificial intelligence. Machine learning is being used for something called “collaborative filtering.”

“Retailers monitor the products all customers view before making a purchase, and recommend similar products to similar customers,” wrote Mannes. “This model has proven very effective for companies like Amazon, Spotify and Netflix. However, the model tends to benefit larger companies over smaller ones. Models improve with data, and it can be a challenge to overcome the cold start problem.”

Propulse isn’t targeting those large companies, as one might suspect, but rather small and medium-sized boutique companies. They charge customers based on marketing improvement. “This requires companies to track changes in conversions, but it’s an easier pill to swallow for retailers that are pitched AI solutions to improve revenue on a daily basis,” wrote Mannes.