My filter for successful visual search is simple – can you take a photo of someone else’s shoes or jacket when on a busy train and find a direct replica online? Can technology negate the awkwardness of actually speaking to someone during your commute to find out where his or her “must-have” item is from?
Fashion stalker claims aside, the short answer is still no. In the majority of cases, the tech is not yet good enough to pull apart a busy image and identify exactly what that piece is.
It is however getting better at finding similar items. Thanks to artificial intelligence, it can identify shape, colour, print and more – it can serve up relevant options and at least start to inspire discovery.
That’s the key theme behind the launch of e-commerce site ASOS’s visual search launch on its native app today.
This is a fashion website with a huge 85,000 products on it; 5,000 new ones every week. One of many challenges in the online retail space is balancing that newness with the overwhelming nature of volume, particularly for users increasingly browsing on mobile. It’s for that same reason we’ve also seen Pinterest and eBay recently playing in this computer vision space. It’s about that keyword: “discovery”.
This rollout from ASOS then, aims to enable shoppers to capture fleeting moments – whether that’s someone they pass on the street, a look a friend is wearing or even a screengrab from Instagram or otherwise – and use them to search through the site’s product lines to find similar suggestions.
“The depth of our offering is absolutely one of our strengths. However that range can be challenging to present to customers, especially on a mobile phone,” Richard Jones, head of product and UX at ASOS, explains to me. “If you know what you want, you can quite simply get to what you’re looking for. But what we’re trying to find is more of that discovery use case – if you’re not quite sure what you want, or you’ve seen something that’s inspired you, visual search is designed to kickstart that discovery… It’s about getting as close as possible to giving you something that is visually similar.”
The tool is shown as a camera icon in the search bar of the ASOS app. Tapping on it then invites customers to either take a picture or upload one from their library to have it find similar products.
Jones knows the tech isn’t yet perfect, if anything the examples out in the market to date have been a “bit clunky”, but with machine learning and big data, it’s only going to improve, he suggests.
ASOS’s own version, the tech for which is powered by an external third party the company has opted not to disclose, is built on this notion. “The more [this tech] gets used, the better it gets trained, the data results get better… the smarter it becomes,” he explains.
That also reflects the way the ASOS team are operating – pushing the launch out to market (in the UK only at first) in order to test and iterate accordingly. It’s about getting it out there and learning how it’s best used before then rolling it to different geographies thereafter.
In its press release, ASOS refers to this as the “build-measure-learn” approach to innovation, a methodology developed by the Lean Startup.
This announcement also supports wider planned technology investment by the company. It currently has a tech team of 900 employees and is planning to hire a further 200 over the next year, for instance. It says its focusing on its AI-powered recommendation engine, which uses big data and a smart algorithm to learn customers’ preferences over time, as well as on improving site infrastructure to drive agility and speed up innovations for customers.
Zooming in on the mobile experience is particularly key. Today 80% of UK traffic for ASOS and nearly 70% of orders come from a mobile device, with people spending 80 minutes per month, on average, in the ASOS app.
With such mobile-native customers, Jones says it’s about how to now use the underlying technology that is in these devices – the high processing power, the ultra high-definition cameras, the depth perception imagery and more.
“We’re thinking about how do we use these devices in a way that is natural and contextual to how our 20-something customers live their lives. They go everywhere with [their smartphones] – so how can we make sure we give them an experience they are expecting?” he comments.
Further motivation lies in the fact using the camera as a means to search is going to become fairly default in September when Apple launches iOS 11, which includes the ARKit development platform. That essentially means all manner of augmented reality uses will be possible directly through the iPhone’s camera lens; visual search included. Net-a-Porter is another e-commerce player that has referenced using it.
“What we want to do is be able to meet that customer expectation and demand,” Jones adds. The visual search tool will live within the app for now, with the intention of making that broader experience an increasingly personalised one for each shopper down the road.
ASOS’s visual search launches on iOS in the UK today with pending rollout for Android and then international scale thereafter.
This post first appeared on Forbes