Sift Chrome Extension
Introducing intentional friction to the online checkout experience
ROLE
Product Designer
TEAM
Just me!
CONTEXT
It started with a closet I couldn't stand to look at.
Four years ago, mid-lockdown, I stared at a closet full of clothes I no longer liked. I upcycled what I could, donated the rest, and felt immense relief.
92 million tonnes of textile waste are produced globally every year. 85% of it ends up in landfills.
The scale is enormous. The root of it is personal. That's where I started.
PROBLEM
Why do we keep buying things we don't need?
First, I wanted to understand the cycle. Why does someone who knows they already have too many clothes, still click buy?
I reviewed the research, then ran a netnography: probes on online forums, responses collected, and a few participants interviewed.
(Claude artifact)
(figjam screenshots)
One pattern kept showing up: People already pause at the cart.
They leave things in their bag for days. They remove an item, then add it back. They close the tab and come back tomorrow.
TARGET AUDIENCE
Who is Sift for anyway?
(Visual)
CURRENT LANDSCAPE
I explored two tools that solved parts of the problem
I spent time looking at existing tools that help people buy less.
The two tools I discovered, chased the same goal, but took very different routes to get there:
→ The Impulse Judge surfaces humorous roasts you have to type out before you can check out. A creative, effective trigger point, but the design is rough and the humor wears thin fast. I tested it with past interviewees.
(image)
→ Think Twice, Buy Once is a mobile app that asks you to wait before buying and tracks the money you save. Clean layout, but there's no contextual pop-up. You have to manually enter every item you're considering, completely outside the moment of shopping. Too much friction, in the wrong place.
(image)
Neither covered: the moment (checkout, in context) and the experience (warm, non-judgmental, designed with care).
CONCEPT TESTING
To test my hunches, I went analog.
One image stuck: what would a well-meaning friend say to someone about to impulse buy? Someone who cares enough to ask, but never guilts you about the answer.
I workshopped those nudges into three lenses, drawn from initial research:
money · space · need
I printed them as card packets and handed them to about 70 people, to pull out whenever they shopped. Analog on purpose: I wanted the stakes real, used in the moment, not in a simulated prototype.
(image)
I went analog on purpose. I wanted the stakes to be real, used in the actual moment of shopping, not in a simulated prototype.
22 people wrote back. About 45% still bought the thing, but all of them paused, and all said they'd cherish the item more for it.
Those responses sharpened the copy, tuned the tone, and gave me the confidence to start building.
THE PRODUCT
So, what is Sift?
A deck of cards between you and the checkout button.
Sift is a Chrome extension that intercepts the checkout button and fans out a deck of playing-card prompts before you buy.
(screen recording)
Design note: The card metaphor and visuals came straight from the physical kit!
The pillars at the center of it all, are the three lenses interviewees kept describing their regret through.
Money. Can I afford it? What am I giving up to have it?
Space. Will it have a home? What has to move to make room?
Need. Do I want it or need it? Am I shopping for my real life, or my fantasy one?
(flow)
(a sentence explaining the flow)
DESIGNS
How Sift works, and what makes Sift, Sift.
→ Pick your pillars. Choose which are active, change them mid-session, and skip at any time. It hands agency back to the shopper instead of taking it away.
(screen recording)
→ Reflect as much or as little as you want. Toggle between the deck and a laid-out view that shows every reflection at once. The laid-out view came from testing: people couldn't see all their notes at once, so I let them spread the deck flat.
(screen recording)
→ The reflections don't disappear. Saved ones land in a quiet history page: what you almost bought, which pillars you weighed, what you told yourself. Less a dashboard, more a journal. The point isn't any one decision, it's noticing your patterns.
(screen recording)
→ Exits everywhere, on purpose. Every exit assumes the best of you, that you mean to cherish what you buy. Even leaving does quite a behavior-change work.
(screen recording)
APPROACH
Making a pause feel grounded and integrated.
Everything leans calm: warm parchment and stone, a terracotta accent, espresso text, each pillar with its own soft accent. It all lives in one Figma design system, so color, type, and components stay consistent.
(DESIGNED IN FIGMA FIRST)
(visuals)
I designed in code and prototyped with AI, directing the build prompt by prompt while I made the design calls: metaphor, motion, copy, exit logic. That let me chase details I'd normally only mock up, the deck fanning out, a reflection tucking toward the toolbar like you're filing it, the small press on every button.
(THEN MOVED TO CLAUDE TO PROTOTYPE)
(a video gallery of micro-interactions)
POTENTIAL IMPACT
What one pause can add up to.
Sift only asks for a couple of minutes, and the minutes aren't the point. A pause becomes a deliberate choice. Choices become a habit. The habit becomes a wardrobe you actually wear, and enough of those, across enough people, chip at the cycle that fills landfills.
(Image of claude artifact)
REFLECTIONS
On the problem
The issue was never awareness. People know they overshop, and that they'll probably regret it. But willpower depletes, and it's up against platforms that are very good at what they do. Adding friction instead of removing it was a new kind of problem to design for, and I'm glad I took it on.
On building with AI
It changed what I could put in front of people. Instead of a static mock, I could direct a real interaction into existence and feel whether the motion carried the calm I wanted. It pushed my role toward decisions, and let the building keep up with the thinking.
NEXT STEPS
Prototype to production.
Next is shipping this as a real extension. The concept is tested and the prototype works end to end; production is what I'm in now. I'm not a coder, so I'm working through how to run Sift cleanly on real retailer sites, and how to keep people's reflections safe.
Further out, I want to bring Sift to mobile, where targeted ads really do their work. That's harder: apps can't overlay each other the way an extension sits on a webpage, so I'd keep the heart of it and rethink the form.
Either way, I'm glad to be starting from a tested concept that actually moves behavior.