How a Tinder for clothes became the basis of Appir Search

Appir started in 2024 as an app where you swiped through clothes. Right if you liked it, left if it wasn't for you. With every swipe, the app got a little better at knowing your taste.

Two years of learning what people like

To make that app work, we had to teach a model what a piece of clothing actually is. A product title like "SS25-SHIRT-04" doesn't tell you much. The model had to learn what something looks like and how it's worn.

About 4,500 people used the app, and together they swiped more than a million times. Every swipe was someone telling us: yes to this, no to that.

Show people random clothes and they like about 9% of them. By the end, they liked about 40% of what our model showed them.

Why we took the app offline

Along the way, we saw where finding clothes really breaks. A few steps earlier, inside the webshop. People arrive with a picture in their head, type it into the search bar, and get nothing. Or, more and more often, a page full of products that look nothing like it.

That's where what we'd built was most useful. Someone searching a webshop usually knows what they want already. They just need to be able to find it.

So we took the app offline and put everything into search for fashion webshops. It wasn't an easy call, because people enjoyed using the app.

Why we do this

A piece of fabric can keep you warm. It can also give you that bit of extra confidence at a party or on your first day at a new job. That only works if it's something you actually like.

A lot of what we buy, we buy because it was shown to us, in an ad or a post from someone we follow. The "social" part of shopping.

But there's one place in a webshop where it runs the other way. In the search bar, the shopper tells you what they want.

"White linen blouse for a summer wedding." "Something yellow with palm leaves for vacation, short sleeves." Those are people with a picture in their head.

And fashion is where search lags behind. Only 39% of fashion revenue comes from search, against 55% across all industries (Prefixbox, 2024). That’s odd, because shoppers who search a clothing webshop buy more than twice as often as those who don’t (13% against 6%, Constructor, 2024).

We want every shopper to be able to find their own style, and the clothes that go with it. That means the search bar has to understand what they mean.

Where we are now

Appir Search is live in the Shopify App Store. You install it from your Shopify admin, with no code and no developer. Results show up in your own theme, so nothing about how your shop looks changes.

It reads whole sentences, including color, fit, fabric, occasion, size, and price. It reads colors from the product photos themselves and fills in the attributes nobody tagged in the catalog. That's how "SS25-SHIRT-04" becomes findable as the mustard vacation shirt it really is.

On another platform? Talk to us, and we'll build the connection.

We also have a demo environment. Send us your shop URL and we'll build a private demo store from your own products. You search your own collection the way your customers do, before you install anything.

What we're working on now is software that audits a webshop's search. Which shopper descriptions dead-end, which ones return the wrong products, and where there's room to improve. When it's ready, you'll read about it here.

What this blog is for

This is the first post here. From now on, we'll write about what changes in Appir Search and what we learn about how people search for clothes. We'll also cover what's happening in fashion e-commerce, and what we think it means for a webshop's search bar.

Expect one or two posts a month.

Curious what Appir Search does with your collection? Send your shop URL to remco@appir.nl and we'll build you a free demo store, no strings attached.