Definitely not what singer Shawn Mendes is wearing. Picture: 123RF/WAVEBREAK MEDIA LTD
Definitely not what singer Shawn Mendes is wearing. Picture: 123RF/WAVEBREAK MEDIA LTD

Amazon has expended plenty of effort to become your go-to place to buy clothes, and it is clearly making inroads. In particular, it appears to be scoring in the category of basics such as undershirts and socks. But the site still isn’t a great shopping experience for more fashion-oriented purchases — and it will need to be if it is to lure more marquee apparel brands and increase its market share.

So I was curious when Amazon recently launched a new feature in its app called StyleSnap. The artificial intelligence-powered tool allows users to upload a photo — a screenshot of a celebrity’s Instagram post, say, or maybe a photo snapped of a friend at a party — and it will offer up similar outfits for purchase on Amazon.

It’s a solid idea that has the potential to address a real pain point many shoppers don’t even quite realise we experience: often, we don’t have the right language to tell a search engine or store worker what we’re looking for. Your idea of “floral print dress” might be a bold garment stamped in big blooms that would fit right in at Miami Beach; someone else’s might be a dainty Liberty-printed frock.

This applies outside of apparel, too; the sofa you think of as “minimalist” might be what furniture retailers categorise as “mid-century modern.”

The great promise of a visual search tool such as StyleSnap is that it could help erase those gaps. And in doing so, it could theoretically give Amazon a unique edge in the e-commerce fashion game. However, I’ve been experimenting with the tool for the past two weeks or so, and I’m not impressed. It got some things right, but generally, I found it to be unhelpful — sometimes comically so.

I also input an image of stylish singing family group Jonas Brothers wearing a flashy, green-textured suit; the first recommended match for his bottoms was a pair of sweatpants

I’ll start by giving credit where it’s due. I uploaded a photo of me, taken last summer at a gelato shop in Italy, and the first dress it offered up was an impressively close match to the one I was wearing, down to the small centre cutout.

But much of the time, StyleSnap missed things that are abundantly obvious to the human eye and brain. I uploaded a screen grab from singer Selena Gomez’s Instagram account. One of StyleSnap’s first four matches for her jean jacket is … a faux fur shawl in Cookie-Monster blue?

The mismatch might reflect that StyleSnap currently only works for dresses, tops and bottoms (something Amazon’s blog post announcing the tool didn’t make clear). But in other instances, when I searched for garments outside of those parameters, StyleSnap gave me no results and a message that I should use it for specific garment categories. Here, it just gave me bad results.

I tried another photo from singer Shawn Mendes’s Instagram feed. He’s wearing a printed, button-down sleeveless shirt. But StyleSnap’s “matches” included grandpa sweater-vests — and one vest that looks ready for a colonial-era costume party.

I also input an image of stylish singing family group Jonas Brothers wearing a flashy, green-textured suit; the first recommended match for his bottoms was a pair of sweatpants. I added a photo of a friend’s pre-school age daughter wearing a pink, orange and red sundress; it returned sundresses for adult women.

I’m not picking on StyleSnap for the sport of it. In part, I’m scrutinising it because Amazon invited us to hold it to a rather high standard by declaring in a blog post that the tool “will change the way you shop, forever”. But, more importantly, it is part of a pattern of apparel offerings and shopping experiences from Amazon that fail to convince me it is going to trample fashion-oriented competitors anytime soon.

Amazon Spark, an Instagram-like shopping tool, is being re-imagined as a new feature called #FoundItOnAmazon, a change that suggests Spark, in its original incarnation, wasn’t exactly a runaway hit. I haven’t seen any data that suggests its Echo Look device, in which voice assistant Alexa offers fashion advice, is having a material impact on its sales in the category. As Shira Ovide and I have noted before, Amazon’s search parameters in apparel are sometimes bewildering — though we have noticed some recent improvements there.

Selling fashion is simply different than selling laundry detergent or cellphone charger cords. Shoppers are often looking for inspiration and curation; they’re in discovery mode, not knock-out-a-quick-errand mode. Amazon seems to get that, and that’s clearly why it is trying things such as Prime Wardrobe, where you can try on clothes before you buy them. But I don’t think Amazon has demonstrated yet that it has cracked the code of discovery-oriented shopping. And so investors ought to be more skeptical about how that dynamic is a hurdle for Amazon in this category.

I’m sure StyleSnap will get better over time — that’s how artificial intelligence-powered tools usually work, after all. But for now it is a glaring reminder that Amazon is on anything but a glide path to decimating its competitors in the fashion arena.

• Halzack is a Bloomberg Opinion columnist covering the consumer and retail industries. She was previously a national retail reporter for the Washington Post. This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.