Not Polyester: From Friction to Feature
Shipped
Most clothing sites let you filter by size, colour, and price, but almost never by what a garment is actually made of. For me that’s often the deciding detail: I want the cotton kurta or the linen shirt, not the polyester one styled to look like it. On the site I shop from most, finding those meant opening every product page one at a time to read the fabric composition buried in the description.
The data was always there (this dress is 100% polyester), just tucked inside “Materials & Wash Care” on each individual page. You’d only find out after clicking in.
The first attempt
I tried to fix this about two years ago. The product idea was clear to me even then: the site already has the fabric data on every page, it just never lets you sort or choose by it. But I couldn’t get from idea to working tool. Building it was beyond what I could put together on my own, and it stalled.
Before: 715 results, and no way to narrow them down by fabric.
The second attempt
I came back to it recently, and this time was different. With AI to build alongside, a couple of prompts got me from “here’s the problem” to a working tool, which was exactly the part that had blocked me for two years.
What it does is simple: pick a fabric, and get back every product made of it. Under the hood it has to know the fabric of every product first, so it reads the composition off each one across the catalog and pulls it into a single tagged, sortable index. A small Chrome extension reads the pages, and a bit of Python collects and organises them. Once that index exists, choosing “linen” and seeing the linen is trivial. I still use it.
After: a fabric bar right where it belongs, so one click narrows 715 results down to just the cotton, or skips the polyester.
What I take from it
The best product features are often the simplest ones. This is a one-line idea: let people choose clothes by fabric. No machine learning, no redesign, no roadmap epic, just data the site already had, made answerable. It’s a small feature, and that’s the point. A site can ship a dozen sophisticated capabilities and still miss the dead-simple filter a real shopper has wanted for years, because it looked too minor to bother with. As a PM I’ve learned to take that instinct seriously and chase the small, obvious fix a user can feel right away before reaching for the impressive one. The simplest thing on the list is often the one worth shipping first.
The gap between my two attempts is the other half of it. Same problem, same idea, same me. What changed is that AI shrank the distance between having a product idea and holding a working version of it. The thing that blocked me for two years wasn’t the idea, it was the build, and that barrier is largely gone now. That’s the shift I care about as a PM. It isn’t AI as one more feature to bolt onto a product, it’s AI changing who gets to build and how fast an idea becomes real. This was me feeling that firsthand, and it’s why I keep building this way.