Pet-commerce is the first-ever online store that uses facial recognition to let dogs do their own shopping. This was only possible thanks to the use of creative data. To build a totally new artificial intelligence, capable of identifying the most varied dogs facial expressions, we had to feed it with twenty thousand images of different dog breeds. Each of them, manually ranked by a dog trainer, according to five levels of interest. After a long process of machine learning, the A.I. was ready - along with the first-ever canine UX – to an unique experience to both dog and owner.

Petz is the largest pet shop chain in Brazil and the company keeps growing. Its e-commerce platform is fast, safe and complete. But brazilians still prefer shopping at one of the eighty stores around the country. So, our challenge was to promote Petz e-commerce - driving traffic to the platform and boosting sales – as well as to show how the company understands what owners feel for their pets and their needs when shopping online.

Pet-commerce is the first online shopping experience designed specifically for dogs. The owner goes to Petz’s website, click on Pet-commerce button and go to an ambient designed specifically for dogs. Then, all the owner has to do is turning on the camera and playing the products videos to the dog. A tool that combines A.I. and visual recognition will analyze each of the dogs facial expressions and detect when he/she is interested in a certain product, adding it to the shopping cart. Then, the owner can review the dog’s wish list and proceed with the purchase.

To teach the artificial intelligence to detect the most varied dogs facial expressions, we fed it with twenty thousand images, from different breeds, manually ranked by a professional, according to five levels of interest. To define these levels, the professional analyzed the position of the ears, mouth, eyes and signs of relaxation or alert. After a long process of machine learning, the A.I was able to mimic the dog trainer ability and detect when the dog insterested or not in a certain product featured on the e-commerce.

Petz’s primary target consists of women (55%) and men (45%), aged 25 to 55 years old, from classes A, B and C. These people treat their dogs like a family member. And they have strong bonds with brands that understand their feelings.

When it comes to innovations and technology, this target is very demanding. They are open to try new things but everything has to work perfectly. So, to offer the best experience for Pet-commerce users, we had to be the most accurate possible. And the key for that was data gathering.

To pull this off, we studied the most common dog breeds in Brazil. Then, we collected images of facial expressions from these breeds. Then, we gathered images of the less common breeds as well as stray dogs, totalling twenty thousand dogs facial expressions.

These images were then analyzed by a professional dog trainer who ranked each of them in one out of five levels of interest. And all this data was used to train a totally new artificial intelligence. After the process of machine learning, the A.I. was able to identify the same levels of interest as the dog trainer did, for every single breed.

Pet-commerce is an unprecedented tool. Facial recognition has never been used to detect dogs levels of interest and excitement. That’s why we had to build a totally new topology of artificial intelligence. And to do that, we had to gather the most complete data information possible. It all started with the collection of twenty thousand images of dogs facial expressions from different breeds. Then, all these images were analyzed by one of the best dog trainers in Brazil, Leonardo Ogata, who took into account the position of the ears, mouth, eyes and signs of relaxation or alert. Then, he ranked each image in five levels of interest: very interested, interested, may be interested, not interested and not looking at the camera. All this data was then used to train the artificial inteligence, through a long process of machine learning. Once the A.I. was ready, we applied it to the first-ever canine UX and integrated it with Petz regular e-commerce platform. The whole process took up to six month to be implemented, from data gathering to launching.

During the first week after lauching, the site traffic increased 236%. The average time spent on the e-commerce jumped 103%. And the sales in toys category lifted 57%, showing that consumers really enjoyed the experience of having their dogs shopping along with them. Furthermore, Pet-commerce earned more than R$3.1 million in spontaneous media. Articles about the project were published in more than 100 media outlets in countries such as Brazil, United States, England, Russia, Australia, Germany, France, Lithuania, Slovakia, Vietnam, Thailand, Japan, Uruguay, Argentina, Georgia etc. And, finally, the positive mentions on social networks reached 95,6%.