Green Ridge · The case study

A better way to choose a campsite, before you get there.

Green Ridge State Forest has 100 first-come campsites, limited cell service, and no reservation system. I built the Green Ridge app to make the trip easier to understand before you leave and easier to navigate once you're there.

Overcast · 46°Flive from the product, as you read thisgreenridge.app

The bottleneck

100 first-come campsites across 49,000 acres, little signal, and one paper sheet at headquarters as the only record of who is camping where.

The system

An offline-first guide campers use to plan a trip. Its record comes from public data and photos of the sheet, and every fact shows its source and its age.

The playbook

Put a working build in front of real conditions early, show what the information doesn’t know, and keep a person on anything uncertain.

A six-panel hand-drawn storyboard of two campers planning a trip, comparing sites, reading the sign-up sheet at headquarters and finding their first choice taken.
The bottleneck: the sheet, drawn first
The public page for Campsite 65 on greenridge.app: the modeled facts (688 ft, southeast-facing, the drive in and its grade, running water 112 ft west), booking, fee and check-in, the line that says the page cannot know whether the site is free tonight, and the map with the nearest sites.
The system: a page that names its gaps
A Green Ridge journey open in Storybook: the persona and scenario on the left, the storyboard frame for step one on the right, and a proposed change attached to a step.
The playbook: a journey in Storybook

The presentation

The deck, as presented: the six-stage trip, what the app knows and what ages, the terrain and the night sky, the sheet and its reader. Arrow keys or swipe to move through it. Open it full screen

The journeys: one record from research to working screens

The design work stays attached to the working product. This is the part I most want another designer to inspect.

I use Storybook for more than component documentation. It is where I can walk through a user journey, see the current interface, collect feedback on the part of the journey where the issue occurs, and work through a change without losing the original context.

The flows on the research board are the same journeys that run in Storybook, one record from research to working screens. This is the Clipboard Upload journey: ten steps from the kitchen table to the sheet at headquarters, drawn panels for the moments off the app and production screens for the moments in it. A research artifact; the drawn panels are illustrations, the screens are captures.

There is a larger reason I work this way. I build with code early because some of the hardest problems in Green Ridge are environmental. Reception disappears. A photograph fails to upload. A map is too heavy for an old device. A site pin turns out to be wrong. A static mockup can help with an interaction, but it cannot tell me whether those conditions have been handled. By keeping the design workspace close to the running product, I can bring actual failures back into the journey instead of documenting an idealized version of how it should work.

02 From feedback to production, inside Storybook · 01 The research and the product direction

I built against the real conditions

Green Ridge started with a map, site information, and an attempt to read the sign-up sheet. It became several connected products.

The first working version went online quickly because I needed to learn whether the idea held up against real handwriting, unreliable signal, and the way campsites are actually distributed across the forest. A polished prototype could not answer those questions.

The app at desktop width with the connection cut partway through, recorded in October 2026: the campsite list, a campsite page and the field guide still open from the device, a banner says the data is cached, and the availability calendar says how old the sheet photo is.

00 The work behind the app · 12 Designing for no cell service · 08 Designing availability when the information gets old

Understanding a campsite before driving there

A dot on a flat map is useful for navigation, but it hides the things that make one campsite very different from another.

Green Ridge’s campsites are spread across ridges, hollows, river terraces, and long gravel roads. A dot on a flat map is useful for navigation, but it hides the things that make one campsite very different from another. I wanted people to understand the ground before they chose where to sleep.

Site 38, from the lines on a map down to the ground under the trees. Aerial: USDA NAIP 2023; the ground view is modeled, not a photograph. Open it in the app

04 Mapping a forest for different kinds of decisions · 05 Rendering the ground and the sky · 11 Making campsites comparable

Uncertainty stays visible

Reading a handwritten sheet without trusting the reader too much.

At the kiosk 0/14 rows · 0%

One photo of the sheet becomes a list of open sites.

It starts at the kiosk. Ahead of the bar it is a photograph; behind it, three columns are rows and two are struck out.

    It runs when this tab opens. The sheet is drawn for this page: every real one carries a column of campers’ surnames, so none is published here, but its structure, the twelve block boundaries and the steps are the pipeline’s own. No model runs in your browser, and the reader takes the twelve blocks in its own order rather than top to bottom.

    DECISIONThe interface can say what appears on the sheet, or that the last record is too old to judge. It keeps the responsibility for registration with the official clipboard and the people at headquarters.
    EVIDENCEI built an admin review interface because confidence scores are not enough. A reviewer can inspect the photographed sheet against the extracted rows, correct individual values, and approve the parts that can safely become part of the shared snapshot.
    The paperparse viewer: a photographed camping roster with every row overlaid, green where a row was accepted and red hatching where it was dropped, beside counts reading 45 accepted, 0 review, 22 dropped, 14 never returned, and a panel explaining that nothing routed to review is not the same as nothing going wrong.
    Nine recorded runs, switchable by capture condition. Green rows were accepted, red hatching was dropped, and the panel on the right says the part that matters: nothing was routed to review, which is not the same as nothing going wrong.

    08 Designing availability when the information gets old · 09 Privacy by design · 10 paperparse and the human review system

    One system: design stays attached to production

    Feedback on a journey step becomes a prototype beside the live page, and a decision.

    A replay of one real run. In Storybook, a reviewer walks the Site Locator journey and leaves feedback on a step. That feedback is filed as a GitHub issue, a Claude worker drafts candidates behind a switch, and the candidates come back into the same step beside the page as it runs in production. The owner picks one, the other is closed as superseded, and the approved change ships.

    Issues 38 and 39 on the site locator, replayed: the step, the feedback, the board, prototype A beside the live page, prototype D chosen and shipped. Read more in chapter 3.

    02 From feedback to production, inside Storybook · 17 Testing, releases, and corrections

    What happened when people actually used it

    The product has to make sense to people who haven’t heard the explanation.

    I can learn a lot from my own trips and conversations with campers, but the product has to work for people who don’t know me. That’s why I pay attention to what people do once they find Green Ridge.

    The milestone I keep coming back to is a stranger taking a photograph of the paper sign-up sheet at headquarters and submitting it through Green Ridge on their own. It’s a small action, but it asks something of the camper. They have to know the feature exists, decide it is worth using, photograph the sheet, and complete the contribution. What they get back isn’t simply a personal feature. They help make the information more useful to other people heading into the forest.

    Search Console exposed a specific kind of intent I hadn’t given its own front door: people were looking for places suitable for meteor showers and viewing the night sky. That wasn’t simply an invitation to write a stargazing article. It was a reason to make sky visibility a campsite-selection criterion and connect the resulting shortlist to the site-level visual experience. The result is the open-sky campsites page and a clearer path from search to a modeled view of the actual skyline from the pitch.

    23 What people actually did · 14 The public website, search, and privacy-conscious analytics · 22 Making the forest findable

    What generalizes beyond this forest

    The Green Ridge app isn’t a business. The pattern travels.

    There’s no revenue, retention curve or customer list to cite here: it’s a free guide to one forest. What it does have is an architecture for a common commercial problem. The real record lives on paper, the work happens where the network drops, and a wrong answer costs someone a wasted trip, or worse.

    UNCERTAIN RECORDSMake provenance visible.
    FIELD WORKAssume connectivity will fail.
    HUMAN REVIEWEscalate ambiguity rather than hide it.
    PHYSICAL OPERATIONSTurn scattered observations into a usable record.

    The short version: a deck of the place, the flows and the decisions →

    Password

    The chapters

    You opened Chapter 18, “Testing, releases, and corrections”. The chapters hold the detail behind this case study and are password protected. Enter the password to continue straight to it, or request access below.