Case study - Mobile · Self-initiated side project

A field companion for plants you have to learn by name

Herbarium is a self-initiated side project: a mobile app that helps horticulture and garden-design students build, document and study a personal plant collection. It’s currently in the prototyping stage - not built, and not yet tested with real users.

My role
Solo product / UX designer - research, IA, UX, UI
Timeline
Started 2026, ongoing - prototyping phase, not yet built or user-tested
Tools
Claude, ChatGPT
Onboarding and home - the daily plant fact as the way back in

Overview

Context · Problem · Constraints

01

Context

Study material scattered across multiple places.

Field identification of trees, shrubs and perennials is a core skill for horticulture and garden-design students, but the way most students actually practise it is fragmented: paper notebooks for field sketches, phone photos with no structure, voice memos that never get written up, and separate reference books or websites for the botanical facts that tie it all together.

By the time exam season arrives, a student's study material is scattered across multiple different places. Herbarium is a personal side project to explore whether a single, purpose-built app could hold the whole loop - capturing a plant in the field, and studying it later - without forcing students into generic note-taking or photo-organising apps that weren't designed for this workflow.

Capture in the field, study at the desk - currently two disconnected habits.

02

The problem

No tool combines fast field capture with structured, citation-backed reference data and a way to actively practise recall.

Students need one place to capture a plant in the field - photo, note or voice note, location - and later study it against reliable botanical facts.

Who's affectedHorticulture and garden-design students building a personal plant reference collection over a semester or a whole course of study.
If unsolvedStudy material stays scattered and inconsistent, recall practice is left to improvised flashcard apps with no domain-specific content, and the effort put into field observations never turns into a structured, reusable, shared reference.

03

Constraints

Named honestly - they're still open.

Reference photographyEvery plant profile needs a small, consistent set - habit, leaf, flower, fruit, bark and the winter/leafless silhouette. Sourcing enough high-quality, consistent photography to seed even a first batch is a significant amount of manual work for a solo project.
Daily fact contentThe home screen's daily plant-fact card needs a sustainable way to generate accurate, engaging short-form content at scale - a concrete pipeline rather than one-off manual writing.
Storage & securityThe database mixes public reference content with users' private notes and field photos, so developer input on data architecture and security is needed before any real build - explicitly outside my expertise, and flagged as a pre-launch step rather than something to design around alone.

Discovery

Research · Synthesis

04

Research

Informal and directional, stated plainly rather than dressed up.

  1. 01My own field observations - identifying and documenting plants outdoors, where the gap showed up firsthand: none of the tools I reached for were built for this loop
  2. 02Field trips - watching how people actually behave on site: hands full, poor signal, capture something fast and structure it later
  3. 03Conversations with garden-design students - informal interviews on how they study for identification exams and where their patchwork of tools breaks down

Notable findings

Photos taken in the field are rarely useful later on their own - without the plant’s name, location and date attached immediately, students lose track of what’s what.
Voice notes are a natural fit for field conditions - hands full, gloves on, walking - but are almost never transcribed or organised afterwards, so they quietly pile up unused.
Students want the option to compare their own field photos against a reliable reference image, rather than relying purely on memory or a textbook photo.
Recall practice - naming a plant from a photo - is improvised with generic flashcard tools that have no botanical content, so students build every deck by hand.

This research is informal and directional, not a formal usability study - appropriate for a side project at the concept stage, and given exactly that much weight.

05

Synthesis

Three needs that had to coexist in one product.

Capture, fast, in the field

Photo, voice and location logging with as little friction as possible.

Structure, after the fact

Turning a loose capture into a proper plant profile, cross-referenced against reliable botanical facts.

Recall, on demand

A way to actively test knowledge, not just re-read notes.

Design principle

Never ask a student to do data-entry work a photo, a voice note or an existing reference source could already do for them.

Concretely: automatic transcription for notes, a shared public plant database instead of per-user redundant entry, and reference facts sourced from vetted publications rather than user-authored text.

Design

Spec first · What changed

06

Spec before screens

The project started as a plain-text functional specification, refined iteratively in conversation with Claude before any visual design began - spelling out entities (plant profiles, families, groups), flows (uploading a list of plants to study, resolving ambiguous name matches, bulk photo recognition) and edge cases, before touching any screens.

Writing the spec this explicitly first - including open questions I hadn't yet resolved myself - made the later design decisions faster.

07

What changed once I prototyped it

Four parts of the brief that moved meaningfully.

Cut
The offline/online distinctionThe original spec called for full offline capture with later sync. Designing the flows showed it added UI complexity - sync states, conflict handling - disproportionate to the value for a first version. Cut for now in favour of a simpler, always-connected model.
Changed
Public database, not per-user dataRather than each student building a private plant list from scratch, the database is public and shared - students browse it and mark plants as favourites. Closer to a shared field guide than a personal spreadsheet.
Added
Public vs. private notesNotes, photos and videos attach to any plant, and each note can stay private or be visible publicly - a lightweight social and reference layer on top of personal study material.
Held firm
Fixed, cited reference contentBase reference photos and descriptive facts are curated and fixed, not user-editable, sourced from three vetted references - Okrasné dřeviny by Martin Vlašák, pladias.cz and the ČZU tree catalog - so the study material stays reliable regardless of who is using the app.

Collaboration here is a solo-designer-plus-AI loop rather than a cross-functional one: Claude for structuring the spec and generating UI, ChatGPT for exploring copy and content ideas. There's no PM, engineer or fellow designer in the loop yet - a real limitation of a side project, and part of why developer expertise comes in before the build phase.

Current state

The core loop, end to end at low-to-mid fidelity

08

Discover & filter

A public database students browse and filter by height, family or group and plant status, with quick category chips - trees, shrubs, perennials, houseplants.

Objevuj - browse the public database, filter by height, preference, soil and group

09

Field capture

A quick-add action - photo or video, voice note, or new plant - sits one tap away from anywhere in the app, backed by a native-feeling camera flow and a photo-upload component.

Quick add - three entry points, and the capture flow behind them

10

Voice notes with transcription

Recorded notes show a waveform, playback and an editable auto-transcript, with a visible nudge to double-check the botanical terms the transcription engine may have got wrong.

Poznámky - private or public, recording in progress, transcript you can correct

11

Locations on a map

Every recorded sighting can be plotted on a map, with a "closest to you" list for planning a field trip around what’s nearby. The map is also available on a plant detail page.

Mapa poloh - nearest sightings first, and the same map inside a profile

12

Plant profile

Each profile combines fixed reference photos and vetted botanical facts with the student’s own field photos, voice notes and written notes - public or private - in one place. Reference photography is organised into the categories that actually matter for identification: habit, leaf, flower, fruit, bark and leafless winter state, as tabs on the profile rather than one undifferentiated gallery.

Plant detail - photo tabs, key facts, and sourced usage notes

13

Collection & practice

Beyond browsing the public database, students search and filter their own saved plants by name - Czech or Latin - or by notes, and jump into a practice hub that tracks progress through flashcard sets by group, like "shrubs flowering in spring", alongside quick links to their collection, plant families and the map of recorded locations.

Zkoumej - flashcards, photo test, and past results

14

The plan

In order, and no further than is honest.

01Finish the remaining screens, iterating on information architecture and navigation, and polish the two study/quiz modes to bring the prototype to full coverage of the specified functionality.
02Test the prototype with a small number of garden-design students before writing any production code.
03Once validated, move into a build phase with an AI-assisted development tool - Lovable, Codex or Claude Code are the current candidates.
04Before that build starts, bring in a developer to review data storage and security, particularly around the split between public reference data and private user content.
05Define a concrete content pipeline for the recurring "fact of the day" feature, and a sourcing plan for the reference photography each plant profile needs.

15

Reflection

Writing the spec out in full was the most valuable stepIncluding the open questions I didn't yet have answers to. It surfaced scope questions - offline support, public vs. private data, where reference photos would come from - much earlier and more cheaply than discovering them mid-prototype.
The biggest open risk isn't the UIIt's content operations and development. A polished plant-profile screen is worth very little without a real pipeline behind it for reference photography and daily facts, and that's the piece I have least solved right now.
Prototyping changed the brief, not just the pixelsOffline sync, shared vs. private data and note visibility all shifted once real screens existed - an argument for prototyping earlier, not for specifying less.
A side project earns its honestyNo users have tested this yet and no code exists. Showing it as a prototype with named gaps is more informative than presenting it as a finished product.

16

Click through it yourself

Herbarium - prototypeOpen in a new tab ↗︎