Many plots, little time
An agent may visit several farms and samples in one day.
Design response
A route-led home screen keeps today’s work and pending tests close.
Strip Test
An agricultural field-testing app UX case study: I designed a mobile workflow for agents to verify plots, capture potato quality samples, and recover offline.
01 · Problem
A quality check is only useful if the field context and the sample are both trustworthy.
Manual potato quality checks across large fields are slow, inconsistent, and prone to capture errors. That uncertainty flows into pricing and yield decisions, so the product needed to make each test easier to perform and easier to trust—without inventing certainty where field conditions make it impossible.
One field agent. Many farmers, plots, and imperfect conditions.
The primary user manages a changing set of farmers and plots, works outdoors, and often arrives with weak signal or uncertain GPS. The dashboard reduces the mental load: what is next, what is pending, and what is safely saved.
- Several farms and plots can be active in one day.
- Gloves, glare and interruptions shape every interaction.
- The agent must know what is saved before leaving the field.
Grounded in direct stakeholder conversations—not a formal research programme.
The direction came from stakeholder conversations that captured notes about current field checks, offline work, and report sharing. I treated those notes as working hypotheses, translated them into design changes, and kept the gaps visible for later validation.
Signal is situational
Coverage can disappear at the exact point of capture.
Design response
Local save states and explicit sync language preserve momentum.
Location needs interpretation
GPS is useful evidence, but field boundaries and accuracy vary.
Design response
Show accuracy, explain uncertainty, and offer a humane retry.
Three rules for decisions in the field
Every interaction had to earn its place against these constraints.
Make the next action obvious
One primary action per state. Supporting information stays quiet until it is useful.
Keep context attached
Plot, sample and result remain visibly connected throughout the workflow.
Never hide uncertainty
Location, save and sync status are explicit so the agent can act with confidence.
Four chapters, one continuous field record
The flow compresses an operational task into a small number of confident hand-offs: identify, verify, capture, review, submit. Captions focus on what each chapter gives the agent, not on the feature inventory.
01
Get started
Start with the right plot and enough context to trust what comes next.
- Open today’s assigned route
- Confirm farmer, field and variety
- See incomplete setup before starting
✳The next task is explicit, while incomplete setup stays visible without blocking route planning.
02
Capture
Guide the sample in context without turning the field task into data entry.
- Verify the agent is inside the plot
- Photograph a consistent sample
- Record only essential observations
✳Capture guidance appears at the moment it is needed, not as a tutorial to remember.
03
Results
Review the field record while the sample and location are still fresh.
- Compare sample outcomes
- Correct an individual entry
- Preview the report before submit
✳Results stay editable until submission and retain their sample context.
04
Offline and recovery
Protect progress when signal drops. Make status legible and the next action credible.
- Save locally without blocking work
- Show what is waiting to sync
- Resume from the last safe step
✳A saved state is more reassuring than a spinner when the network disappears.
Recovery is part of the core flow
Field tools are judged at their edges. These states make the problem legible, preserve work where possible, and give one credible next step.
Weak GPS
Show current accuracy and let the agent retry from a better position.
Outside plot
Explain the boundary clearly before asking the agent to move.
No signal
Continue safely and label what will sync later.
Photo issue
Keep the sample and ask only for the image that needs replacing.
Submit failed
Preserve the report and offer one credible retry action.
Verify the agent is inside the plot before a test begins.
That gate protects the integrity of the record, but it adds friction at exactly the moment an agent wants to move. The answer was not to remove the check; it was to make uncertainty understandable and recovery humane.
- The location check protects the provenance of every sample.
- Accuracy is shown as a range, never as false precision.
- When verification fails, the agent keeps context and receives one next step.
✳
The principle: protect the record without punishing the person doing the work.
Warm enough for field work. Strict enough for status.
A restrained russet palette carries actions and hierarchy; Noto Sans keeps the interface plainspoken and readable. Green is reserved only for good results and confirmed success.
Russet leads action and focus. Cream and ink carry structure. Green communicates confirmed success only.
Plainspoken labels, generous line height, and strong numeric states remain readable outdoors.
A coherent field record from plot verification to report.
The product enables an agent to identify the right field context, capture consistent samples, preserve work through connectivity gaps, and leave with a traceable report. This is a designed capability—not a claim of measured impact.
What the design enables
A clear path from assigned plot to a saved, traceable field record.
What remains unknown
Speed, completion rate and error reduction still require field validation.
What I’d test next
GPS recovery language, sample guidance and hand-off into report sharing.
Evidence boundary
This case study describes designed capability—not measured impact.
Screen overview
Depth lives in the transitions, not only the happy path.
A secondary map of the Section 1 screen set, recreated with identifying content removed.