Second-hand fashion online booms

How AI in search and discovery offers new opportunities

The Corona pandemic has brought many changes to our lives. We have been grounded in our homes and our lives revolve around a much smaller orbit that usually extends from fridge to table. People are discovering new hobbies, and investing in home sports equipment or finally getting a pet. This also means economising on space at home. People have also started to get rid of old stuff in a manner not seen before. I remember reading headlines which suggested that non-profit organisations are asking people to stop donating old clothes as they simply couldn't handle any more stock.

A winner

It’s no secret that online second-hand fashion stores have flourished because of these changes in these circumstances. And what's not to like about the idea that you can simply upload a picture of an old sweater and get rid of it while making a bit of cash.

It sounds easy and it is. If you wanted to sell something on Vinted for example, all you need to do is upload several pictures of the item, tell them what type of item it is, the name of the brand and in what condition the product was in – and suddenly you're an online seller. In addition, the fashion industry has long been under scrutiny for its lack of progress in curbing its environmental impact.  

Lyst’s annual Year in Fashion report, which brings together tonnes of data on the most popular brands, products, people and movements in the past 12 months, also reported a rising interest in used clothes. According to the report the term “vintage fashion” generated more than 35,000 new searches.

This has also brought to light new business models like online vintage search engine, Gem, which specialises in purely vintage and secondhand fashion. Clearly the opportunity has been spotted, and what is interesting to see is the range of prices and brands on offer – from high street, to vintage to luxury – consumers are looking for a bargain or simply a unique product.

 

The problem

However, as this industry is on the rise, it also presents a problem that "regular" online retailers already struggle with, and for second-hand retailers the gap widens even more.

The challenge? Good quality data.

The consumer at home who is selling her or his leopard printed legging collection and therefore suddenly sees them self as a fashion expert is not necessarily a copywriter or a marketing guru (if you actually know of someone that has a leopard print legging collection and is a marketing guru, please let us know we might have a position for her/him).

Instead, we end up with a catalogue full of products that have very little information attached to them. This makes product discovery for any shopper very difficult - without the right tags the search results are vague at best.

Here are some examples of where you can see how “poor” the data is most of the times:

We see the same problem here:

 

Or even on ebay:

 

Or at ASOS Marketplace they have been savvier, but in general there is a brand and short description:

 

So there are clearly there are many opportunities to buy second-hand, but quite simply if a shopper can’t find what they are looking for they will go elsewhere. This is a problem that all online brands have to counter, but in the case of second-hand product items, there are often no clear collections or product ranges to filter products down to what a shopper is looking for.  Instead, consumers are likely to find the equivalent of a cluttered online bargain bin!

 

How can AI help?

The good news is that breakthroughs in AI for commerce have made it possible to improve user experience.

At Attraqt, we’ve implemented various AI tactics such as machine learning, deep learning, information retrieval, natural language processing and computer vision to identify and extract all types of information or signals that describe a product and a search query.

Traditionally, products tend to be described or tagged with textual data. However, at Attraqt we are able to tag visually with product images, attributes, as well as data on how users have interacted with that product.  For instance, searches can be paired with data about products that have been clicked or purchased after the search.

This way we are able to create what we call a "product fingerprint” even if there was very little information attached to it, and the results we have seen so far are staggering.

- CASE STUDY

Pretty Little Thing

 

DOWNLOAD THE CASE STUDY

And it gets better

This is all groundbreaking, but in addition, AI in search also allows us to simultaneously deliver visual recommendations.  With this we automatically recommend products that look similar, even if they are not described in the same way. What’s more, the AI then automatically recommends products that are similar to the style the model wears in the shot. (When I say model, I mean the person with the leopard leggings collection) As you can imagine it isn’t easy to pull off a good “complete look” wearing leopard leggings, but if you do, you’d want to be able to recommend other products similar to the leopard print design or leggings, or a complete outfit. We call these Shop the Look recommendations.

 

 

The possibilities are endless, and what this offers every brand, online trader, budding designer or the next second-hand fashion mogul is the opportunity to make their products easily searchable, and at scale.

I’m really keen to follow the progress of sites like Gem, and Vintage who have made the most of the impact of the pandemic where they have turned second-hand into a trendy proposition to inspire discovery and new shopper experiences.

 

CONTACT US

You may also like

Ready to see our platform in action?