Why force consumers to scan barcodes, QR codes or, even worse, make them search for product numbers? Wouldn’t it be a better buying experience just to point your smartphone at a picture of the product you want to buy and have it instantly loaded? That’s exactly what Notonthehighstreet have done with a new feature for Christmas 2019 seeign their entire Christmas catalogue underpinned by AI and image recognition, making it instantly shoppable through the notonthehighstreet mobile app (on iOS).
notonthehighstreet app users simply hold their phone over an image in the Christmas catalogue, which then instantly finds the product and makes it available for purchase directly through the app.
“We wanted to better connect our customers offline and online experience. Discoverability is a key development area for us, so this feature is just one way we’re making it easier for our customers to shop with NOTHS. We know that people love browsing our catalogue, and many do so while also searching our site through their mobile, this new feature removes the need to hunt down product codes and brings people the perfect thoughtful gift instantly.”
– Richard Zubrik, Chief Technology Officer, notonthehighstreet
The user experience of the feature is deceptively straightforward. However, a combination of human effort and software-driven image augmentation resulted in a dataset of over 1 million images of the 331 products from the 100 page catalogue. This was required to train the image classification deep learning model, which took over 168 hours. The feature was built in-house at notonthehighstreet using Apple’s image classification frameworks (CoreML and MLImageClassifierBuilder).
“More than 70% of traffic to our platform comes via mobile devices, so we are constantly looking at how we can improve our mobile user experience. AI and image recognition technology are both opening up new opportunities for brands like ours, which are reliant on visual data, to make our mobile offering more responsive to our customers and adaptive to their shopping behaviour.
This feature of our mobile app is a first test for us and something we will be looking to build on in the New Year as we strive to create a far more seamless shopping experience, helping our customers to find the perfect thoughtful product for all occasions.”
– Richard Zubrik, Chief Technology Officer, notonthehighstreet
4 Responses
ummmmmm…. that’s sort of sound in theory, image search never really been my cup of tea tbh, but i don’t think someone’s followed the logic from NOTHS.
people generally search for things to buy that they don’t already own, if they already own them they don’t need another of the same thing (barring consumables, which we really don’t need image search for).
when ebay introduced image search, i was cynical, still am. their approach was along the lines of “like your mates look? steal it” which was not very cool to say the least.
but if you want a specific pair of adidas sneakers, you don’t know the name or model code, then taking a picture is a good solution. in theory.
apply this to “not on the high street”, where their entire gimmick is it’s not on the high street, or anywhere else, you need to get it from NOTHS….. so where are you supposed to be taking the picture from? on the high street? it’s not there apparently.
and am i reading right that it only matches to 133 products? what’s the point in that? a list of 133 items is not too long to browse through, why not just say “hey, want one of these 133 items on this list? no? don’t bother with image search then”.
This sounds good for the sellers who manage to get selected and featured but what about the 99% of sellers who aren’t in the catalogue? NOTHS should fix their very broken search engine first so that some of the lovely products on the site can be discovered in ordinary ways.