The ecommerce value of AI image recognition, as proven by pugs & poodles

We asked ourselves, how might we use AI to inject more emotion into the purchasing journey?  This led to us challenging our developers to embrace machine learning in a new and innovative way –  with a pet project (in more ways than one!) focused on image recognition.

Can AI help enhance emotion in ecommerce?

Emotion is a powerful part of ecommerce. Imagine an owner of a chihuahua puppy visiting an online pet supplies retailer. When they land on the site, they’re thinking about buying a product their dog will enjoy: an emotion-led task. But having to navigate through the full range of dog food, much of which is irrelevant or unsuitable, makes that emotion fade. The task becomes an entirely functional one – correlating with a reduced basket value and lower likelihood of the customer returning.

Blockers to personalisation

By collecting some basic details, the pet store could serve up personalised content so customers only see relevant products and their emotion stays high. Stores usually gather this user data as a customer interacts with the website – personalisation increases with the amount of data. But ideally, we'd have more information on the first visit, as great initial impressions are crucial in winning loyal customers.

Until now, getting that data has required a sign-up form – a laborious interaction guaranteed to turn a feel-good purchase into a tedious one. We're back to square one.

From chore to standout experience

By extending the applications of AI, we can collect the information in a better way.

A photo could remove the need for customers to manually input information. In our pet food example, an image could reveal a dog's breed, age, size and – with the introduction of video – even health issues. This turns a boring-but-necessary process into an interaction that strengthens the emotional connection between you and your customer from day one.

Proving the ecommerce potential

To see if image recognition could provide the data retailers need, we set our developers a challenge: create an AI-powered app that can recognise nine common dog breeds with high accuracy – from a picture alone. 

The pet example was carefully chosen – dogs are famously problematic when it comes to image recognition. From chihuahuas that resemble muffins to pugs that look like loaves of bread, the nuanced nature of dogs’ appearance confuses computers. If our team could achieve consistently accurate results using dogs, it would act as proof of concept for the potential of image recognition in almost any industry.

Proof of concept in just 5 days

Our team configured the AI platform and sourced and categorised 20,000 training images. They manually QA-ed hundreds of results, refined the process countless times and tackled the various challenges of mixed breed dogs and distracting backgrounds. After just 5 days, the system was returning correct responses for all nine breeds with a high level of accuracy.

Increasing conversions with an emotional connection

Not only does image recognition allow you to gather essential information in a customer-friendly way, it also provides you with a personalised image for every customer. Including this personalised imagery into content blocks, banners, checkout flow, profile icon, emails and more adds further emotion into the mix.

This pet project did more than just demonstrate the power of AI, it showed that looking for new applications of existing technology should be on every major retailer’s roadmap. And proved the speed at which we can develop a fun, engaging, production-ready AI.

Want a demo of Pet AI?

Get in touch