In a world obsessed with optimizing everything—from your grocery list to your hinge profile—it’s tempting to think that AI could just “solve” fashion. Plug in your preferences, your budget, your body type... boom. Wardrobe, complete.

But fashion isn’t a spreadsheet. It’s not a problem to solve—it’s a story to tell.

And that’s why GPTs (and other AI search tools) fall short. They’re built to filter by function. Not to spark emotion.

Let’s break it down using a little quadrant magic (because don’t we all miss high school math). Let’s put ‘Cheap’ to ‘expensive’ on the Y-axis and ‘selling function’ vs. ‘selling emotion’ on the X-axis. Shoutout to Curious Commerce who initially introduced this concept and whose post ‘Can AI be my personal stylist?’ really influenced my thinking here. Definitely worth the read!

In one corner, we’ve got cheap and functional—think Amazon Basics, Costco socks, the kind of stuff you buy while reordering toothpaste. This is automation at its peak. It’s not stylish, it’s not emotional. It’s efficient. Agents are going to take over this space and I’m here for it.

Then there’s cheap and emotional—this is the domain of fast fashion. TikTok trends. Zara hauls. Shein’s 42 tabs open. It’s driven by attention, FOMO, and instant gratification. Feels good for five minutes. Leaves you with the fashion equivalent of a sugar crash.

Next up, expensive and functional. Think bridesmaid dresses, hiking boots, performance gear you research for hours. This is “solve-the-equation” territory. You outsource it to ChatGPT or Google Shopping and pray for a good return policy.

But the top right corner? That’s where the magic lives.

Expensive and emotional.
This is where great brands win. Not because they’re the cheapest. Not because they “solve” your needs. But because they invite you into a world. They make you feel seen, stylish, and smart. They build connection—and that’s what keeps you coming back.

Brands like Bandit Running are doing this masterfully. On paper, they sell performance wear. But in practice, they’re selling identity. Community. Belonging. Their gear signals a mindset, not just a mile pace—and their brand is laced with storytelling, aspiration, and cultural context. You don’t buy Bandit just to run—you buy it to be part of something.

On the luxury end, take The Row. They’ve built an empire not on logos or loud design, but on quiet confidence and whisper-level aspiration. Their pieces are eye-wateringly expensive—and still, people covet them. Because The Row isn’t selling cashmere. They’re selling a lifestyle: restraint, elegance, timelessness. It’s the fashion version of a deep breath. GPTs can’t parse that.

If you’re a brand, this should sound alarm bells. Because the tools that are rising to dominate discovery—AI search, sponsored content, functional queries—are pointing shoppers toward the bottom-left of this map. Efficient. Low-margin. Emotionless.

And that’s not a game most brands can win.

The opportunity? Build for the top right. Build emotional connection. Build trust. Build curation.

That’s where Beni comes in.
We’re not building an AI that tells you what to buy.
We’re building an AI that helps you discover what feels right—for your style, your values, your life.

We clean up the chaos of resale, yes. But more than that, we aim to reintroduce feeling into search. Because we believe the future of fashion needs more connection, not just more options.

In a world where everyone’s optimizing for clicks, we’re optimizing for curiosity. For taste. For trust.

GPTs are the new search engines—but they’re not stylists. They can recommend five brands under $100. But they can’t tell you which one will make you feel like a million bucks. And they definitely can’t tell you which piece you’ll still be reaching for in five years.

Because style is a feeling. Not a filter.

What do you think? Let me know in the comments!

Here’s a great visual example of the kinds of information that are relevant to us:

Our product catalog is similarly structured to how you might organize your closet - we have all of our pants in one “drawer”, all of our shirts in another, etc. When you are shopping for pants, and therefore we are looking for matching pants (but secondhand and for half the price), we don’t want to be searching through our whole catalog of 200M+ listings to find you that exact item because it would take us way too long (and you probably won’t be waiting around). Instead, we want to go to our ‘pants drawer’ and search there specifically. Therefore, determining the category of the product that you are looking at is quite important for us - when you’re shopping for pants, we don’t want to be searching through where we store the shirts, and showing you shirts.

In most cases, we can figure out pretty easily that you’re looking at pants by performing a keyword search with our bank of keywords defining pants (pants, trousers, jeans, etc.) against the text descriptors that we can find on the page (breadcrumbs, product title, etc.). In other cases, the text descriptors that are provided are not actually super helpful in determining the product category.

Here’s an example. On QVC lives a product called the Bernardo Hooded Quilted Puffer Walker. The product page looks like this (also - checkout that 50% off steal!):

In this case, we use a generative AI model, namely GPT, to help us extract additional product descriptors from the information that we have available. What is GPT? You might have heard of it in the context of ChatGPT, but it’s essentially a Large Language Model (LLM) that can generate new, coherent, and contextually relevant text based on the inputs it has received. It’s pre-trained on a large corpus of text data, encompassing diverse topics and domains, which enables the model to learn general structure and nuance of human language. 

Here’s how it works. We feed it the information we do have (you can see it’s not super useful): 

1. Product Title: Bernardo Hooded Quilted Puffer Walker

2. Breadcrumbs: Product Detail | A566808

And we ask it to return some additional product tags, in which case, it says:

• Apparel

• Outerwear

• Jackets

• Puffer Jackets

• Bernardo Hooded Quilted Puffer Walker

Aha! Now we have a lot more information about what the potential product category might be, and we can go searching through our catalog of jackets. And that’s how we use GPT to help us figure out what you’re shopping for with Beni!

Okay, now we’re ready.

Making resale accessible is more complicated than it seems.  With over 200M listings in our database, aggregated from over 40 different resale platforms, we needed a way to cross check the shopper’s wish with our database to deliver a relevant resale alternative in just a few seconds. Moreso, we needed a way to understand the shoppers intent to determine what the most relevant alternative was for that specific shopper.

Enter: Beni’s Ranking Algorithm.   

First, let’s talk about how ranking algorithms work.   

A ranking algorithm is a method used to rank items in a list, based on certain criteria or metrics, in a way that is meaningful and useful for the particular application. They can be relatively simple by using basic statistical methods, or highly complex utilizing machine learning and other advanced techniques. They’re critical for search engines like Google to deliver you the most relevant web pages on the internet based on your search, or recommendation systems built into applications like Spotify or Amazon, sorting items (songs, products, etc) based on user preferences, previous interactions, similarities between items, and other user behavior. They’re important for any application with a “feed” (think Facebook, Twitter, LinkedIn) with a goal of optimizing engagement and relevant content for the user, and they’re important for Beni, with a goal of delivering the most relevant secondhand versions of what you are shopping for. 

So how did we build our ranking algorithm? When we set off, we generated a set of rules that made sense based on our business and what we are trying to achieve, and began by just hard-coding those rules into our application. We match images of secondhand listings in our database to the image of the item you are shopping for, and by doing so, we generate a “similarity score”. Along the same lines, we generate a set of scores from matching on various text-based attributes like product title and brand. We then add weights to each of those elements, along with some other product filterability, to achieve a rank-ordering of products that we feel is most relevant to the user. The algorithm produces a score on the confidence in our ability to deliver an “exact item”, and we deliver to you those resale listings from highest to lowest confidence.

We initially focused on developing logic to rank items with attributes similar to the product you're shopping for, as this was our primary use case for the extension. However, in the broader context of e-commerce, and other areas of opportunity for Beni, different forms of logic can be utilized to generate recommendations to you in different shopping modes, including:

1. Showing you items that are often bought in conjunction with the product you're shopping for.

2. Showing you items purchased by users who share demographic or behavioral similarities with you.

3. Showing you products that possess attributes akin to the item you are interested in buying.

These multifaceted approaches to product ranking and recommendation help us create a personalized and “smart” shopping experience for you, seamlessly connecting you with items that align with your interests and purchasing habits (and ultimately, giving you a sensation that we’ve read your mind!!).

The first questions that we asked ourselves when building our ranking algorithm from scratch were:

• What is the desirable behavior of the system, and when is it achieved?

• What are the metrics that we can measure ourselves against to determine whether we are achieving the desired user behavior?

If our goal at Beni is to deliver the most relevant secondhand versions of what you are shopping for, we know we are doing a great job if you are clicking on our listings and ultimately making a purchase. We can now use this information to optimize our ranking algorithm and test changes. As you might imagine, there is more happening at the “top of the funnel” (e.g. the number of times you expand the extension) compared to the “bottom of the funnel” (e.g. that time when you actually bought something that Beni showed you).

Therefore, to be able to iterate quickly, we optimize for earlier signs of engagement instead of optimizing for our end-state goal which is, of course, finding you that really amazing secondhand alternative for half the price. At this point our ranking algorithm is learning from users like you, by iterating on the algorithm and seeing how it improves your engagement with our extension. Ideally we get to a state where we can get multiple changes out every week. Are you seeing the improvements?! 

When we feel like we’ve got it just right (we’re not there yet, but I’m sure you can see that we’re working hard!), we can begin to train a model to learn from the product rankings that we’ve been delivering to our users.

In some ways, the rules that we developed early on and over time become features to the ranking model. Eventually, the model will be able to learn and iterate on user behavior faster than our development team. When we get there, I promise to share our learnings with y’all once again. Stay tuned!

In order to deliver you the best resale options, there are 3 key areas that our engineering team is focused on:

1. Become experts in normalizing and categorizing the resale data that we get from our partner feeds and extracting product attributes from it that you, as a shopper, care about.

2. Become the smart shopping tool that accompanies you along your online shopping session and that can accurately identify the products you are shopping for.

3. Be able to search through our catalog of 200M+ listings to deliver you the most relevant secondhand versions of what you are looking for in a matter of seconds. 

In all three areas, we leverage the power of artificial intelligence (AI) to achieve these goals. In this post, I'll provide a high-level overview of how and when we’re using AI, and in the upcoming series, I'll delve into specific examples within each area. So, let's dive in!

Normalizing and categorizing resale data.

One of the challenges, and opportunities, with resale is that the product data is super messy.  Imagine a brick-and-mortar thrift store: every single item is one-of-one and there is no structured product info, like a SKU number, to help keep track of things.  Now multiply this by millions and you have a sense of Beni’s database of resale listings.  For example, we often find differences in the way that products are categorized amongst all of our resale partners, especially as it relates to product category, gender, color, brand, and size. If we are unable to appropriately classify products into the right bins, it makes it almost impossible to surface these listings to you in the right situations. For each of the product attributes mentioned above, we use a diverse range of technologies to re-classify items into distinct bins, carefully defined here at Beni.

In some cases, just re-classifying the standard product attributes is not enough to surface a more appropriate listing above another. The Beni catalog has gotten so large, that there are some products out there that we just have literally hundreds of relevant matches for (cue the Bottega Veneta Cassette Padded Leather Crossbody Bag). At Beni, we’re working hard to improve our ranking algorithm to better curate the selection that we deliver to you. There are plenty of opportunities to continue to improve the algorithm using generative AI (AI that generates content, like GPT) to extract product attributes that shoppers care about from the semantics of the data that is present in our catalog. Some examples might include classifying brands into different tiers and only showcasing products from the same brand tier, further evaluating the condition of the item, or refining the potential discount that you might achieve by purchasing the item secondhand.

Accompanying you along your online shopping sessions.

At Beni, we can shop with you on over 1500+ merchant retail sites (and counting). We have developed the technology in-house that allows us to determine whether you are shopping on Patagonia or Reformation, and whether you are shopping for the Nano Puff Jacket or the Frankie Silk Dress. As you might imagine, the structures of these websites have a tendency to change (have you ever revisited a link to an item that you saved on your favorite retail site that no longer exists?). At Beni, we have developed a technology that notifies us when our recipe to read a merchant site is “broken”, and are developing ways to automate a fix using generative AI. 

One exciting recent development is that we can use this same tooling to not only fix existing merchant sites, but also to automatically add support for new merchants on the fly. If our goal is to be the best-in-class at delivering your search in resale, we don’t want to interrupt your shopping experience when you try to use Beni on a site we don’t yet support. Instead, we want to automatically add support for the site you’re shopping on at the time you are actively engaged in shopping. 

Finally, there is a lot that goes on behind the scenes in categorizing the products that you are looking at in real time. Our product catalog is similarly structured to how you might organize your closet - you put all your pants in one drawer, all your socks in another, etc. When you are looking for pants, you don’t go sweeping through your whole wardrobe looking for pants - instead, you go directly to the drawer where the pants are stored. Well, that means that when Beni is shopping with you online, we need to know that you, too, are looking for pants. In many cases, we can do this with a keyword search against a bank of keywords that we have used to define “pants” (trousers, jeans, …). But in some cases, the product attributes that are available to us are not as descriptive as we would like them to be (cue the Bernardo Hooded Quilted Puffer Walker, you can see that this is a jacket, but it doesn’t actually say “jacket” anywhere on the product page). In those cases, we use generative AI to help us further tag the products so that we can appropriately identify them at the beginning of our search.

Delivering search in a matter of seconds.

There’s a lot that goes on behind the scenes when it comes to searching a catalog of resale items, and little tweaks that we make to the algorithm can lead to pretty large impacts in the ranking of the results. At Beni, we are currently using a weighted combination of searching with the product image (using AI to perform the reverse image search), product-related descriptors like the title and the brand (using AI to perform text similarity search), the price, and slicing the data by the aforementioned product attributes (category, size, color, etc). But what excites me most about the future is that with all the recent advancements that have been happening with generative AI, there are increasingly accessible opportunities for Beni to layer technology that allows us to query the data in our catalog based on the semantics or the meaning of the data rather than just querying the data itself, enabling us to satisfy queries that are based on seasonality or trends (things like “find me dresses that are perfect for summer weddings requiring garden party attire”, or “find me pink heels I can wear to go watch the Barbie movie”). 

At Beni, our passion for delivering you an exceptional search experience through the world of resale is the driving force behind our every endeavor, and utilizing powerful technology like AI to deliver those secondhand deals to everyone, everywhere on the internet is at the core of everything we do.

Join us on this series in sharing with you how we utilize our technology to shape the future of resale. 

Have questions you want us to answer in our next post? Comment below!