The world of fashion and trends has been turned on its head by technology in the last decade – and the next big change is about to hit the mainstream.
The first evolution started when consumers began to spend more and more time on social media, resulting in less time listening to the designers, tastemakers and magazine editors who were previously the guardians of what was considered fashionable. Our collective gaze is now increasingly focused on the endless content we stream to our digital devices. We’re increasingly connecting with a new generation of online influencers, who generate much higher levels of trust and engagement with consumers than the traditional media brands of the past do.
This leap forward in technology and content has led to an explosion in the amount of data that’s collected every day – and this is a trend that’s only set to continue. It’s this huge increase in data – and our newfound ability to rapidly make sense of it with artificial intelligence (AI) – that is driving the second evolution in the fashion industry. This potent combination of data and insights now allows businesses to innovate – and offer superior experiences to consumers, especially online.
Robin Wong is the CEO of custom.cm – a fully automated platform that uses machine learning and AI, to help people discover and shop the most exciting trends from influencers on social media platforms. I heard him speaking at the FashHack London investor summit.
AL: Hi Robin. Can you start off by telling me a bit about CUSTOM?
RW: CUSTOM is designed for the social media generation, to make it easy to see and shop the latest trends from across the social media spectrum.
We’ve been working closely with our advisers to provide an experience that makes use of artificial intelligence – not only to find the hottest trends from social media, but also to find live product matches from online retailers that are available for purchase right now. This cuts out the need to spend hours tracking something down once you’ve found it, as CUSTOM automatically finds matches and alternatives.
By collaborating with experts in their respective fields, we’ve created an experience that works hard for consumers trying to find inspiration, for digital influencers who create the content, and for the retailers who want to sell more products. We’re lucky to have input from Susie Lau, who is one of the world’s most famous fashion bloggers; Steve Vranakis, who is executive creative director at Google Creative Lab and president of the D&AD; and Mark Chalmers, who is executive creative director at i-D magazine.
Together, we’ve created a new kind of business. It makes extensive use of automation and artificial intelligence, to take the effort out of discovering and shopping for the latest trends from around the world.
AL: What’s the experience like, for users?
RW: Typically, a user’s first experience of CUSTOM will be in the form of a sponsored post in their social media feed – allowing them to see the latest live trends from around the world, or to shop a trending post from their favourite influencers, or perhaps to see the styles that influencers are talking about right now.
We’re targeting consumers with content very carefully, to ensure we’re being as hyper-targeted as possible. This approach is reaping dividends in the response rates we’re seeing. In fact, consumers are liking the content they see in our adverts so much that almost 1,000 people a week decide to follow us because of them!
Once consumers click through to our website (depending on the type of sponsored post they click on) they can browse posts from fashion trendsetters, to see what people are talking about; see and shop for product matches, or alternatives; or search to find something specific, and see which products are being recommended by influencers right now.
We’ve launched the platform with a website, to ensure that there is the widest access possible – and so that there are as few barriers as possible to experiencing the content we provide. There are no apps to download to get started, no screenshots to upload, and no waiting to see product matches. It’s just about seeing the very latest trends from stylish people on social media – and automatically being able to shop what they’re talking about, right then and there.
AL: Can’t people do this already?
RW: There’s a massive amount of fashion-related content being created around the clock by online influencers. When we analysed the social media channels of the world’s most prolific influencers, we found that more than 95% of posts had no product links – and no simple clues that would allow people to find these products.
Quite often, people were posting tantalising clues like “check out my cool #gucci jacket”, which usually results in a trail of comments, from people wanting to know where an item came from. We thought it was a huge missed opportunity to satisfy that demand – but then we also realised that it’s also a big ask for influencers to put links to everything they take a picture of. Linking is a time-consuming process, and not practical to complete for every post.
Posts with links are a tiny minority – and those products often sell out rapidly, leaving consumers with few options but to hunt around if they want to try and find a match. The current state of affairs means hassle for both influencers and consumers.
Our aim then, is to make all these inspiring posts easy to shop. We also want to provide people with a range of alternatives, in case they can’t afford the £1,000 Balenciaga boots they had seen so-and-so wearing. Inspiration’s a great thing, but inspiration you can do something with is even better.
Our platform also works in a complimentary manner with all parties involved. We don’t touch existing affiliate links, which allows consumers to continue to use those. Brand sponsors have the potential to make more sales from the increased exposure – both on social media, and on the platform. Alternative brands have their products placed in front of consumers at the right time and the right place. Finally, consumers get to see more live trends, which they can shop on the spot.
AL: How does CUSTOM actually go about finding trends and matches?
RW: We knew that by offering people the best posts from the most stylish influencers we could give people something to inspire them – and others in the industry would struggle to do this as quickly as we could. The problem then was about working out what is trending right now, and how to find product matches quickly.
To solve that, we got hold as much social media and product data as possible, and used automation and machine learning to make sense of it as quickly as possible. Firstly, we monitor every social media channel of hundreds of the world’s top online influencers. Then, we’re also monitoring 70 of the best and biggest online retailers to check what they have in stock. That’s around 3.5 million live products that we’re keeping close tabs on – a figure that’s going up all the time.
We use this encyclopaedic knowledge what’s going on in the fashion world, and run it through a system that has been trained to understand how humans and retailers talk about fashion on a very deep level. Our analysis includes everything from hashtags and emojis, to product codes and features. This is the system that automatically finds matches for posts – it allows us to do, in a matter of minutes, something that would previously have taken expensive teams of researchers days to complete.
This system allows us to work out what people are talking about in their posts, and to find product matches for trending posts in several different ways. We even can find product matches where links to products aren’t available, and there is very little information available. We’re about to start combining visual information into the mix, which will allow us to find even more matches.
AL: How many people does the system take to run?
RW: It’s taken 18 months to get to this point. Right now, to keep the system ticking over, and training the machine learning algorithms to recognise new vocabulary and features, it takes about three people. You need some fashion and some tech knowledge to train it correctly, but a big part of it is the data you feed into the system in the first place – and how you structure and manage this.
AL: How does the business make money?
RW: Every time someone buys a product through a featured retailer, we get paid an affiliate commission. When you balance that against how cheap it is to run the business, compared to publishing and retailing, it’s obvious how much more efficiently we can get consumers from the point of discovery to the point of purchase.
We’ve also developed an integration for Shopify merchants. This allows us to plug directly into their stock and order systems. We have several interested parties, who are looking to amplify the impact of social media activity. For this enhanced value, we would charge a premium on sales commissions.
In the future, we are already envisaging offering real-time data intelligence, to both influencers and brands. This would be based on the trove of data we already have, and it could work on a subscription model basis.
AL: Presumably this technology has wider usages?
RW: Absolutely, currently the system is trained to understand the worlds of fashion, jewellery, makeup, footwear and accessories – amongst others. But it can be trained to understand anything that social media trendsetters talk about – like holidays, or other high-value purchases.
As long as we can plug in data where someone is talking about something, and then plug in the retailers or merchants that sell what they are talking about, we can train the system to find a match or a suitable alternative.
AL: Finally, what’s next for CUSTOM?
RW: We’re soon to close our SyndicateRoom crowdfunding campaign – so we’re pushing that as hard as possible right now.
Please let us know your views in the comments below.
- When big data crushes big egos
- Can we love robots?
- Robots are the ultimate job-stealing “imiagrunts”
Category: Artificial Intelligence