A self-initiated concept: a mobile guide that tells a farmer what to do today, from planning and pests to prices and schemes. Prototype tested with 15 farmers, and now being rebuilt.
At a glance
The problem
Farmers keep about a third of the retail price, and pests and disease take an estimated 20-30% of yield. The people who most need guidance are the hardest to reach: only 29% of rural Indians were online.
What I designed
A mobile guide that tells a farmer what to do today, from planning and pests to prices and schemes, built for low literacy, weak signal and local languages.
The evidence behind it
Eleven products reviewed, three proto-personas, and a prototype tested with 15 farmers across Karnataka.
My role
Sole product designer: research, strategy, UX, UI, design system, prototype and testing.
Three decisions that shaped it
How I know it works
Weekly farmers completing their tasks
Read it
01
A self-initiated concept: a farming guide that tells a farmer what to do today, designed for people with a basic Android phone, patchy signal and little time to learn an app.
02
Secondary research showed where farmers lose value, and how hard the people who need guidance most are to reach.
33%
of the retail price reaches the farmer
For fruits and vegetables. The rest goes to middlemen, transport and retail.
20-30%
of yield lost to pests and disease
Warnings and advice arrive too late, or not at all.
29%
of rural Indians were online
Phones are common, but connectivity and digital skills are not.
03
I reviewed eleven products and platforms in the space. They are strong in places, but they mostly start somewhere other than the farmer’s day.
Each dot is a product. Left to right shows how much of the crop cycle the product helps with. Bottom to top shows who it is built for. The green corner is the space nobody covers.
Helps farmers, but only after the harvest
Helps new farmers through the whole season
After the harvest, built for businesses
The whole season, built for businesses
Advice and crop health
Satellite insights, enterprise analytics and advisory. Some are built for agri-businesses, some are text or SMS only.
Inputs and commerce
Strong on delivering seeds and inputs. The advice is usually tied to a sale.
Community
Peer advice and vernacular Q&A. Useful, but not structured around a crop cycle.
Markets and finance
Transparent prices, kiosks, warehousing and finance. Little help before the harvest.
The gap I saw
04
From the numbers to a tested prototype in four steps, each one building on the last.
Secondary research
Industry reports, government data and academic studies, to find where farmers lose value.
Competitor review
Eleven products across advice, inputs, community, markets and finance.
Hypotheses
Three proto-personas, each with a problem, goal and hypothesis to test.
Early prototype and test
A clickable prototype, tested in person with 15 farmers across Karnataka.
05
These are proto-personas, built from research before I met a single tester. They span the range of access and confidence, and testing was how I checked them.
Ramesh, 45
The experienced optimizer
About five acres and twenty years of farming. Uses WhatsApp and YouTube. He knows the land, but weather and prices are less predictable than they used to be.
Hypothesis. If Ramesh gets timely, specific advice and sale guidance, he will rely on the app to decide on his yield and his price.
Kavita, 30
The new farmer
Two to three acres, inherited. Educated and keen, but overwhelmed by choices: which crop, which input, what to do next.
Hypothesis. If Kavita can compare crops and follow guided steps, she will gain the confidence to complete her first season.
Sujatha, 60
The traditional smallholder
One acre, farmed with her husband. Cannot read English and avoids apps for fear of pressing the wrong button. Wants to be told, shown and reassured.
Hypothesis. If Sujatha can see and hear the guidance in her own language, she will use the app on her own.
06
I tested my first clickable prototype with 15 farmers. It was the most useful part of the project: five findings, and a clear list of what had to change.
What I heard
Usable, but language was a barrier
“I can’t read English, but I can follow if someone tells me.”
What it meant
The interface needs voice and pictures, not just better words.
What I changed
Voice narration and picture-led tasks added to the design plan.
What I heard
Too much text confused new farmers
“There are so many options… which is best?”
What it meant
Too many choices up front stalls a first-time user.
What I changed
Crop planning cut to three steps, with comparison cards and filters.
What I heard
Farmers wanted it to work offline
“Network is poor in our field. The app should work offline.”
What it meant
Connectivity is a design constraint, not an edge case.
What I changed
Offline-first plan: cache tasks and the weather forecast.
What I heard
The assistant was barely used
What it meant
People did not know what to ask an open-ended bot.
What I changed
Make FarmBuddy proactive: suggested questions, and awareness of the screen the farmer is on.
What I heard
Selling prices were confusing
“Is ₹2,000 a good price or not?”
What it meant
A number means nothing without something to compare it with.
What I changed
Added mandi benchmarks, a price forecast and sale tips.
07
Counts from 15 moderated sessions. I use counts, not percentages, because that is what a sample this size supports.
12 / 15
followed the stage-wise tasks unaided
The other three needed small prompts, mostly because smartphones were new to them.
13 / 15
felt more confident choosing a crop
After using crop planning, 87% said they felt more confident about what to plant next.
10 / 15
said the price forecast gave them more control
Over when and where to sell.
13 / 15
would use it weekly, if it had voice support
The condition matters: every one asked for support in their own language.
“I can’t read English, but I can follow if someone tells me.”
“There are so many options… which is best?”
09
Five rules that came out of the testing and the research. Each one is something I would hold the final design to.
Tell me what to do today
The strongest reaction in testing was to “what should I do today?”. Lead with the next action, not a dashboard.
Show less first
Too many options stalled new farmers. Offer a few good choices, with a way to see more.
Pictures and sound before words
Language was the biggest barrier. Use icons and, where possible, voice, so reading is optional.
Work without a signal
Connectivity in the field is poor. The tasks and the forecast must work offline, and say how fresh they are.
Give every number something to compare
A price alone meant nothing (“Is ₹2,000 a good price or not?”). Show the benchmark, the range and the trend.
10
Five decisions shaped the product. Each one answers something the research or the testing showed.
01
Lead with what to do today
Options I weighed
Why
The strongest reaction in testing was to “what should I do today?”. 12 of 15 testers followed the stage-wise tasks on their own. Clarity of action beats volume of information.
02
Plan for voice and pictures from the start
Options I weighed
Why
Language was the biggest barrier. “I can’t read English, but I can follow if someone tells me.” 13 of 15 would use it weekly if it had voice support.
03
Work offline
Options I weighed
Why
Signal in the field is poor. “Network is poor in our field. The app should work offline.”
04
Fewer choices when planning a crop
Options I weighed
Why
New farmers stalled on too many options: “which is best?”. A short comparison keeps the decision with the farmer.
05
Launch with five features, not nine
Options I weighed
Why
Pest detection, direct buyers and the scheme explorer each carry risk: a wrong diagnosis can cost a crop, and buyers need trust. They follow once the basics are trusted.
11
Nine features, grouped by the job a farmer is doing. They all sit behind one idea: tell me what to do today. The scope stays the same in the final design, and the experience around it is rebuilt.
Plan
Daily snapshot
Weather, active crops, task alerts and market rates on one screen, so “what should I focus on today?” is always answered.
AI crop planning
Suggests crops from land data, water availability and the forecast, to reduce trial and error from the first season.
Grow
Task-based crop management
Stage-wise checklists for watering, fertilising and crop care, with reminders.
Weather alerts and irrigation tips
Weather turned into a decision: when to water, when to protect.
Pest and disease detection
Upload a photo of a crop and get a likely diagnosis and next step.
Ask
FarmBuddy, the assistant
Part of the app from the start: a chat assistant for questions in natural language, so farmers can ask instead of searching through menus.
Sell and fund
Price forecasting
Current mandi prices beside trends, to choose when and where to sell.
Direct buyer access
Buyer interest, chat-based negotiation and sale planning.
Scheme and loan explorer
Checks eligibility for government schemes and the Kisan Credit Card.
12
Nine features is too many for a first version. This is my prioritisation, based on what the testing showed and on what is risky to build.
Daily snapshot and tasks
The core. 12 of 15 testers followed the stage-wise tasks on their own, and the daily view is what drew the strongest reactions.
Crop planning, simplified
Cut to three steps. 13 of 15 said it made them more confident about what to plant.
Weather alerts and irrigation tips
Cheap to build with existing data, and it turns a forecast into a decision.
Price forecast with mandi benchmarks
10 of 15 said it gave them more control over when and where to sell, as long as the numbers have something to compare against.
FarmBuddy, made proactive
Already in the app. Suggested questions fix the “I did not know what to ask” problem.
13
Three features are valuable but risky. I would hold them back until the basics are trusted.
Pest and disease detection
A wrong diagnosis can cost a farmer a crop. It needs a tested model and a clear human fallback first.
Direct buyer access
Needs verified buyers, payment trust and support. Testers were wary of selling and paying through a new app.
Scheme and loan explorer
Testers valued it, but it works best with direct links to government systems, which is a longer build.
14
An app that advises farmers can do harm if it is confidently wrong. These are the risks I design for.
The risk
The crop doctor names the wrong disease
What I would do
Show how sure it is, give the safest remedy first, and offer a human check for anything serious.
The risk
FarmBuddy gives a wrong or made-up answer
What I would do
Answer only from vetted farming sources, say where the answer comes from, and hand over to a person when it is not sure.
The risk
A forecast or price is out of date, offline
What I would do
Always show when it was last updated, and give a range, not a single number.
The risk
A farmer trusts an alert too much
What I would do
Pair every alert with the reason and a plain next step, and let the farmer mark it as done or wrong.
15
Measures I would track, chosen to show whether farmers act on the guidance, not how many people installed the app.
Weekly farmers completing tasks
The main measure: the share of active farmers who complete their stage-wise tasks each week.
First session to first task
How many new farmers finish onboarding and complete a task without help.
Return rate
How many come back the next week, and the one after.
Decisions made with the app
Crops chosen with the planner, and sales timed with the forecast.
FarmBuddy that resolves a question
Questions answered without needing a person, and how often it hands over.
16
These came straight from the farmers who tested it.
Testers asked for
Offline mode
On the roadmap
Cache the content for low-connectivity areas.
Testers asked for
Voice navigation
On the roadmap
Voice prompts in Kannada, Hindi and Tamil.
Testers asked for
Video tutorials
On the roadmap
Short regional-language explainers for crops and finance.
Testers asked for
Manual chat support
On the roadmap
A layer of live help for questions FarmBuddy hands over.
Testers asked for
Multi-language onboarding
On the roadmap
Language choice in the first screens and in settings.
17
The same flow and the same features, designed again from the ground up: a design system built for low literacy and low bandwidth, new screens and a new prototype. This section fills in as I finish it.
Design principles
Tell me what to do today. Show less first. Speak and show. Work without signal. Earn trust before asking for it.
In progress
Design system
Tokens, components, icon set, accessibility rules and an audio pattern.
In progress
Final screens
Every flow, redesigned, in English and Kannada.
In progress
Interactive prototype
Embedded here, so you can use it.
In progress
18
The screens I tested. The final design keeps the features and rebuilds the UX and UI around what the testing showed.
Daily snapshot
AI crop planning
Task checklist
Weather alerts
Pest and disease
FarmBuddy
Price forecast
Direct buyers
Scheme explorer
19
The biggest lesson was not about farming. It was about which assumptions to test first, and what to design for before you test.
Language is the first screen
I tested in English with people who mostly do not read it. The result was predictable, and it is the first thing the final design fixes.
Direction beats information
The strongest reaction was to “what should I do today?”, not to any dashboard. Clarity of action is worth more than volume of data.
Test the riskiest thing early
Voice came from the farmers, and it was the one thing my prototype could not test. In the final design I plan for it from the start.
There is more behind this
13 more sections, about 8 minutes of reading.
