GearSmart — Research & Design Case Study

Role & Context

Sole designer and builder

Role

Sole designer and builder — user research, product design, schema design, and AI-assisted engineering

Context

Senior capstone, 2026

Tools

Claude (AI gear-list generation via tool use), Vercel, OpenWeather, Open-Meteo, Google Maps/Places, SerpAPI, Figma

Section 01

Beginner outdoor gear
shopping is broken

Every year, thousands of people head outdoors for the first time — and most show up underprepared. Not because they didn't research. Because the information they found was built for someone who already knows.

REI's gear pages assume you know what a sleeping bag's temperature rating means for your elevation in October. Reddit threads assume you know which questions to ask. YouTube reviews assume you have $800 to spend. So beginners overbuy what they don't need, underbuy what they do, or give up and wing it. All three lead to bad trips.

Three root causes sit underneath it:

Pain Point 01

Condition Gaps

Generic gear lists don't account for where you're actually going. A "hiking checklist" won't tell you desert nights in October drop to 35°F, or that a Pacific Northwest trail in June makes rain gear non-negotiable. Beginners don't know what they don't know — and the internet doesn't tell them until it's too late.

Pain Point 02

Cost Intimidation

Walking into REI as a beginner is overwhelming. $600 tents. $200 sleeping bags. Reviews that never mention a budget. First-timers overbuy because they can't tell which items matter most — or they check out entirely because the entry price feels impossible.

Pain Point 03

Information Overload

There's no shortage of gear content. There's a shortage of gear content relevant to a specific beginner on a specific trip. Assembling an answer for your exact situation takes hours and still leaves gaps.

Section 02

The research —
Ladies Climbing Coalition

To move past assumptions, I ran a structured research session with the Ladies Climbing Coalition — a community of women climbers at every level, from complete beginners to multi-year veterans. Real beginners, not classmates: people who remember the confusion, the wrong gear, and the moments where better information would have changed everything.

Group

Ladies Climbing Coalition

Format

2 activities, 1 session

Methods

REI Roleplay · Affinity Board

Activity 01

The REI Roleplay

Each participant drew one of six scenario cards and played an experienced REI employee advising a beginner about a specific trip. They had to think through: what would a beginner most likely forget? What would actually hurt them? What's the first thing you'd hand them? The exercise forced people to externalize expert knowledge they'd internalized — and showed exactly which gaps were universal versus trip-specific.

Scenario 01

Desert camping weekend, first time sleeping outdoors

Scenario 02

Winter day hike, Pacific Northwest, November

Scenario 03

3-day backpacking trip, Sierra Nevada, June

Scenario 04

Beginner rock climbing, Joshua Tree, spring

Scenario 05

Family car camping, budget under $200 total

Scenario 06

Solo overnight hike, total beginner, no gear owned

Activity 02

The Affinity Board

Afterward, each person wrote individual observations and "I wish I'd known" moments on sticky notes — one idea per note — and we grouped them as a room, letting themes surface from the bottom up. Three clusters formed almost immediately: condition surprises (weather and terrain caught people off guard), gear paralysis (too many options, no starting point), and specificity gaps (generic advice didn't match the real trip). Those three clusters became the foundation for everything GearSmart does.

Testing GearSmart with real people from the community. — **Back to the community, tested with real people.** GearSmart started with community research and returned to it — testing with real people kept the project honest about who it's for.

Section 03

Three insights that
changed the design

Not synthesized summaries — the words participants actually used. Each one pointed to a direct design response.

I had no idea desert nights were freezing — I showed up with one layer.

Insight — Condition Gaps

Generic lists don't warn you about what's counterintuitive. Destination and month data has to surface hidden conditions, not just gear categories.

I wished someone gave me a list specific to where I was actually going, not just hiking in general.

Insight — Specificity Gap

"Hiking gear" isn't a useful category. Lists have to be built around the exact trip — terrain, season, duration, activity — not a generic archetype.

I didn't know what was actually essential vs. nice to have. Everything looked equally important.

Insight — Priority Blindness

Flat gear lists create anxiety, not clarity. Beginners need a clear signal for what keeps them safe versus what's an upgrade, so they know where to spend first.

The deeper pattern. Running three target users through the early product surfaced one thing all of them shared: a gap between how skilled they thought they were and how skilled the trip actually required. The early version took people's self-rating at face value. The real problem wasn't gear — it was calibration. That insight became the spine of the next version.

Section 04

Every decision came
from the research

How might we digitally recreate the experienced friend who knows what's needed — for a specific beginner on a specific trip?

GearSmart isn't a gear database with filters. Every structural decision maps back to what participants told me they needed.

The Brief

What we set out to build

Recreate the experienced friend who tells a beginner what to pack — generate a trip-specific list where every item carries a reason, grounded in the real conditions for their destination, dates, activity, experience level, and budget. Validate it with real beginners, and ship it as a live product.

GearSmart trip view showing a weather alert and trip-condition notes from live data. — **Built on live conditions (real weather, alerts, and trip notes).** Every list is grounded in the actual conditions for that exact place and date: climate normals via Open-Meteo (ERA5) and current weather via OpenWeather. It surfaces a plain-language weather alert plus trip-condition notes on elevation, terrain, and UV. This is the data layer that makes the advice trip-specific.

Addresses — Condition Gaps + Specificity Gap

Why AI?

No static database can account for every combination of destination, month, activity, experience level, and budget. AI generates a list that's genuinely specific to each trip instead of a closest-match approximation. Say "Joshua Tree, October, camping" and the model knows desert nights are cold and recommends accordingly — the conversation a good REI employee would have, automated.

Addresses — Priority Blindness

Why checklists with Essential badges?

Participants described real anxiety from flat lists where everything felt equally important. "Essential" badges create a clear hierarchy — what you can't leave behind, separate from what's a nice upgrade. A progress bar turns packing from a stressful task into something completable. Both patterns came straight from the affinity clusters.

Addresses — Specificity Gap

Why a why for every item?

Each item expands to explain why it's on your list for this trip. The reasoning is the product — it's what turns a checklist into advice.

Addresses — Cost Intimidation

Why retailer-agnostic links?

Knowing what to buy is half the problem; knowing where to start — and trusting the price — is the other half. GearSmart routes each item to whichever trusted retailer sells it cheapest, rather than locking to one store. (The story behind that decision is in the next section.)

Tap any item to see why it's on your list for this trip — plus alternatives at different price points.

Section 05

Building it for
real users

GearSmart is a live, deployed product, not a mockup. You type your trip; the site quietly asks three services — a weather feed, an AI advisor, and a price finder — and hands back a list built for your trip, with a reason for every item and a link to buy each at the best price.

The Pivot

The pivot that defines the product

I applied to REI's affiliate program to monetize the product links. REI rejected the application. Instead of tying the product to one store, I rebuilt the entire purchase layer to be retailer-agnostic: it searches across trusted outdoor retailers — REI, Backcountry, Patagonia, Black Diamond, The North Face, Moosejaw, Evo — and routes you to whoever sells each item cheapest. The rejection made the product better.

"GearSmart isn't tied to one retailer's inventory — we recommend the right gear and send you to whoever sells it."

The path wasn't clean, either: a key shopping API was deprecated mid-build, forcing a migration to a different engine to keep direct-to-retailer links working.

Tap to buy and GearSmart routes you to whichever trusted retailer sells that item cheapest — never locked to one store.

Hardening

Making it not-broken

Shipping to real users meant hardening it. The first version leaked API keys in the browser and broke when the AI returned malformed data. I moved every secret server-side, rotated the exposed keys, replaced fragile text parsing with the AI's structured tool-use so invalid output is now impossible, added a graceful retry so users see "trying again" instead of an error, and set hard spending caps. The judgment behind those choices — not the speed of building — is the design.

Architecture: Browser → Vercel functions → weather, AI gear list, and price routing → back to a single trip-specific list. The map stays in the browser, locked to this site.

GearSmart gear checklist organized by category with Essential and Recommended flags. — **The gear checklist organized by category, check as you pack.** GearSmart turns a trip into a finishable checklist with each item flagged Essential or Recommended and a local pro tip for the area.

Section 06

What I designed,
tested, and removed

The strongest design decisions were the ones I walked back. Three features I built, tested, and pulled:

Walk-back 01

Per-person pricing

Designed: multiply gear cost by group size in the cart.
Learned: a four-person list showing $1,200+ per person reads as terrifying, and it muddied the single-user cart.
Replaced with: a quiet "shared" tag on group gear, and a cart that stays personal.

Walk-back 02

Hover tooltips for the "why"

Designed: hover to reveal why an item's on your list.
Tested: it clipped on the first item, and hover doesn't translate to touch.
Replaced with: a tap-to-expand row that carries the reasoning — which became one of the product's best features.

Walk-back 03

A separate solo-mode toggle

Designed: a dedicated solo switch.
Learned: it was redundant.
Replaced with: solo simply derives from a group size of one. One less control.

Each removal made the product simpler. Showing them is the point — design maturity is knowing what to cut.

Section 07

Tested with
real people

Real beginners. The Ladies Climbing Coalition sessions shaped the product before I built it; a later round of remote testing with beginners shaped the refinements. One tester described GearSmart as "a knowledgeable, generous friend." Rhys Newman's note that "beginner / intermediate / expert is too vague" is what drove the calibration work.

Industry reviewers. A panel of working designers reviewed it: a senior product designer at Patagonia (who started her career as a gear-store sales associate — literally the friend GearSmart digitizes), a designer on Adobe's Spectrum design system, and a human-factors product researcher. The Patagonia reviewer called the carabiner-and-light-bulb logo "ingenious" and the budget step "really smart, because outdoor gear is really expensive," and handed me a product roadmap: show pack weight as you pack, add a sustainability rating for the conscious consumer, and build a borrowing community. The Adobe reviewer pushed on visual restraint and on making the human design judgment visible behind an AI-built prototype — a note I've carried into this case study.

The public. At the senior exhibition, visitors walked a four-table flow — "Four Tables, One Conversation" — pulling a scenario card, packing a kit from their own guesses, then typing the same trip into GearSmart and watching a researched list build beside their guess. The gap between guessing and knowing, made physical.

GearSmart exhibition build, Station 02 "What would you bring?" with a gear table. — **An immersive exhibition with a full branded build, station by station.** I built the capstone out as a physical, branded space so the digital tool had a real-world counterpart. Station 02, "What would you bring?", invites visitors to pack from the table on instinct first, before seeing what GearSmart would recommend.

Guiding first-timers through the gear table under the "Pack right. Get outside." banner. — **Simulating the friend, live — guiding first-timers through the gear.** I walked first-timers through the gear table under the "Pack right. Get outside." banner. Watching people reason out loud about what to bring was the clearest test of whether the concept lands.

It's genius that you started pulling the API from weather and from the vendors to bring everything into one shop.

— Sam Anvari, advisor

Section 08

Where GearSmart
goes next

The current version proves the concept. The roadmap is shaped by what testers and reviewers asked for.

Now

Live and deployed: trip-specific lists, calibrated experience levels, group and solo trips, retailer-agnostic price routing, and an AR Fit Guide that coaches you through fitting a backpack correctly once it's on the list.
Soon

A running pack-weight total beside the price total, a °C/°F toggle, durable caching and rate limiting, and forecast-based weather so a trip months out reflects the conditions you'll actually face.
Later

A sustainability rating built on verifiable signals, not a made-up score — surfacing real certifications and take-back programs so the conscious consumer can choose well. Pack weight that changes which pack gets recommended. Expanded activities beyond hiking, camping, backpacking, climbing, and skiing.
The vision

Saved lists and return trips, a post-trip feedback loop so the tool learns what you actually used, and a borrowing community so the most responsible option — borrow before you buy — is built into the product.
AR Fit Guide

Getting the gear is only half of it. GearSmart's AR Fit Guide walks you through fitting a backpack correctly — hip belt, sternum strap, load lifters, shoulder fit, torso length — so the pack you bought actually works on your body.

Step 1 — torso measurement — The AR Fit Guide reads 17 body keypoints live, in-browser, and coaches you through five fitting steps — then confirms the fit and remembers your torso size for next time.

Step 2 — set the hip belt — The AR Fit Guide reads 17 body keypoints live, in-browser, and coaches you through five fitting steps — then confirms the fit and remembers your torso size for next time.

Reflection

The tool is the artifact.
The judgment is the design

GearSmart taught me what AI is good for as a design material: speed. It let me build and test in days instead of weeks. But the defensible work is the judgment underneath — framing the real problem (calibration, not gear), designing the reasoning structure, the retailer-agnostic pivot, and testing with real beginners instead of classmates. The tool is the artifact. The judgment is the design.

Why we built
GearSmart

Sole designer and builder