A true farmer's companion, direct market access, fair pricing, and AI crop health for rural India.
The result, first
I designed 7 core features for low-literacy rural farmers, replaced one overwhelming registration form with a 5-step conversation, and removed the middleman from the farmer-to-buyer deal flow entirely.
Product
KrishiConnect, a mobile platform giving rural Indian farmers direct market access, live pricing, AI crop-health checks, and government-scheme guidance in their own language.
Business problem
Farmers lose an estimated 15–20% of their commodity value to middlemen because they lack market information and direct buyer access.
Target users
Rural smallholder farmers with limited formal education, low digital literacy, 2G connectivity, a regional language, and often no prior smartphone experience.
Rural farmers in India are basically selling blind. Without knowing the real-time commodity prices, they end up accepting whatever the middleman offers, which is often only a fraction of the actual market rate.
KrishiConnect connects farmers directly with buyers, shows them live market data, and puts expert farming guidance into a hand that may have never held a smartphone before. The main constraint throughout was low digital literacy, so every flow, label and step had to work for a farmer sitting in a village on a weak connection.
This was a self-initiated concept, so there was no client, no PM and no engineering partner. That meant the discipline to validate my own assumptions had to come completely from me.
I lead with the 7 features and the 5-step onboarding because for this user the measure of good design is whether they can finish a task at all, not how refined it looks.
Who I was designing for
With no stakeholders to interview, I depended on secondary research to define the problem before designing anything, mainly academic work on rural digital literacy, NABARD reports, and a competitive study of AgriApp and Dehaat.
The most useful shift in my thinking came early. The middleman is not really the villain here. The real problem is the information gap. Farmers described the agents as a natural result of not knowing the market, so the goal became giving farmers the same information, and not making them suspicious of agents.
About
Grows a small plot and sells each harvest through the one agent within reach, with no way to know the day’s real market rate.
Goals
Get a fair price for the harvest
Reach buyers directly
Read prices without doing maths
Frustrations
No visibility into the market rate
Dependent on a single agent
Text-heavy apps built for cities
Read a rising trend line instantly but struggled with raw numbers, so visual encoding had to carry the price meaning.
About
Buys produce directly from farmers and needs to verify sellers and check precise price history before committing to a deal.
Goals
Source produce directly
Verify who’s selling
See exact price history
Frustrations
Hard to confirm a seller’s identity
Farmer-first views hide detail
Precise data buried or missing
Wanted precise historical price tables, a real need I served with a toggle without cluttering the farmer’s default view.
Before KrishiConnect, a farmer with a ready harvest and no idea of the market rate would call the one agent he could reach, hear a price he had no way to verify against the mandi, and sell at whatever was offered, losing 15 to 20% he never even sees. This “selling blind” journey is exactly what the whole design set out to break.
1Harvest ready
Ramesh opens KrishiConnect with a crop to sell.
AnimationMarket snapshot loads
Check price
2Read the trend
Live price shown as a rising or falling graph, not raw numbers.
AnimationTrend line draws in
Find a buyer
3Reach a buyer
He connects directly with an Aadhar-verified buyer.
AnimationVerified badge appears
Make a deal
4Sell direct
He negotiates and sells, no middleman, a fair price.
AnimationDeal confirmed
Ramesh, Smallholder Farmer
Wants a fair price for his harvest without depending on a middleman he can't verify. This current-state map of "selling blind" framed the whole design.
Harvest
Find a Buyer
Negotiate
Sell
Crop is ready to sell
No sense of the current market rate
Calls the local agent
No direct line to real buyers
Agent quotes a price
No way to check it against the mandi
Accepts the offer
Loses 15–20% to the middleman
Satisfaction
“Good crop this season.”
“The agent is the only person I can reach.”
“Is this fair? I have no way to know.”
“I took what I could get.”
I built the strategy around the farmer’s own words, “the agent is the only person I can reach”. That one sentence, and not a feature list, is what defined the actual problem the product had to solve.
What made the cut, and what didn't
A self-initiated project has no PM to say no, so I had to impose that discipline on myself. Every feature had to justify its place as if a stakeholder was watching, which kept the set down to seven features a farmer could actually learn.
The clearest scope call was the price display. I kept the farmer view as a simple trend graph and put the exact numeric tables behind a toggle for buyers, instead of compromising both.
Shipped in v1
Home screen: market, weather, and crop health at a glance
Market Trends: live commodity prices as a readable graph
Community: direct connection to Aadhar-verified buyers
AI Crop Health Monitor with a shareable PDF report
Regional-language support and government-scheme guidance
Deliberately cut
Numeric price tables as the default: Demoted to a buyer-only toggle; raw numbers failed the low-numeracy farmer they were meant to serve.
Offline fallback for AI crop health: Not designed in v1, the gap I most regret, since the feature degraded badly on 2G (see Evaluation).
Deep buyer-side tooling: The buyer persona was deliberately under-scoped to protect focus on the farmer, a trade I’d rebalance next time.
Content strategy did two jobs: translate every label into the farmer’s regional language, and turn government-scheme information into plain guidance a farmer could act on. Trust was content too, Aadhar-verified badges on buyer profiles gave a government-backed signal no visual could fake.
Keeping the feature count to seven was the whole game here. One extra “nice to have” feature just means a farmer who gives up on the app before he even reaches the one that actually pays his bills.
How the product was organised
The structure had to make sense to someone who reads a screen by its shape before its words. So I kept it to a small set of top-level tabs, each one doing a single obvious job, so that navigation never depended on reading a label correctly.
Onboarding was the real centrepiece of the structure. A single multi-field form was getting abandoned instantly, so I broke it down into a step-by-step flow with one question per screen.
KrishiConnect
Onboarding
Language select
Aadhaar verify
Crop profile
Home
Market snapshot
Weather
Crop-health shortcut
Market Trends
Price graph
Farmer trend view
Buyer table (toggle)
Commodity list
Community
Verified buyers
Negotiate / deal
My listings
Crop Health
Photograph crop
AI assessment
Shareable PDF
Schemes
Government schemes
Eligibility guide
Onboard, one question per screen, 5 steps
Instead of a wall of fields, each screen asks a single question and advances. It adds two screens over a one-page form, a deliberate length-for-comprehension trade that dropped abandonment to near zero in testing.
Check a price and reach a buyer
The farmer opens Market Trends, reads the direction from the trend line without interpreting numbers, then moves to Community to deal directly with an Aadhar-verified buyer. The branch is farmer versus buyer: buyers can toggle a precise table the farmer never has to see.
Diagnose a crop
The farmer photographs a crop and receives a plain-language health assessment plus a PDF to share with an agronomist or buyer. In low connectivity this flow degraded badly, the one place the branching should have included an offline path and didn’t.
I designed the onboarding like a conversation instead of a form, because the drop-off was happening at the very first point of friction. So removing that wall of fields mattered much more than reducing the number of steps.
From lo-fi thinking to hi-fi execution
Every wireframe had to answer the same question. Can a farmer tell what this screen does before reading a single word? If a layout needed its label to be understood, it had failed.
With no engineering partner, the prototypes had to carry the validation weight on their own, since they were the only way for me to test the low-literacy bet with proxy users.
The low-fidelity layouts leaned on icon, position, and imagery to signal function, with text as confirmation rather than the primary cue. I tested whether a screen’s purpose survived with the labels blurred out.
In high fidelity, the pivotal change came from a single round-1 observation: farmers read a trend graph instantly but stalled on numeric tables. That killed the numeric default and made the graph the farmer-facing display, the most impactful decision in the project.
I stress-tested the layouts with the labels blurred out, because for this user a screen that only works after you read it does not really work at all.
The look, the system, and the words
The surface is built to be read by shape and colour first. Live prices show up as trend lines, the actions are large and icon-led, and the visual language stays consistent so that a pattern learned on one screen carries over to the next.
Trust is also a visible design element here. The Aadhar-verified badges make credibility something a farmer can actually see at a glance, before he commits to a deal.
Trend-graph price display
Encodes direction visually, so low-numeracy users read “up or down” without interpreting numbers.
Icon-led action patterns
Function communicated by shape and position first, text second, reusable across every screen.
Aadhar trust badge
A consistent, government-backed credibility marker on buyer profiles.
Onboarding question“What do you grow?”: One plain question per screen turned a form into a conversation and removed the cognitive wall.
Verified-buyer badge“Aadhar verified”: Two words carrying a government-backed trust signal no visual flourish could replace.
Crop-health result“Leaf looks healthy. Tap to save a report.”: Plain-language reassurance plus a single next step, written for someone who won’t parse clinical terms.
The microcopy had to survive translation and low literacy at once, so every string was written short, literal, and free of anything a farmer would have to decode.
What testing and the numbers said
I tested with demographic-proxy participants, people from a rural background, with a regional-language preference and limited smartphone experience. This validated the graphical price display and the step-by-step onboarding quite early and quite clearly. The decision to show a graph instead of numbers came straight out of a round-1 observation, which is the kind of insight desk research would never have given me.
7Core features designed for low-literacy rural users
5Onboarding steps replacing one long overwhelming form
0Middlemen in the farmer-to-buyer deal flow
AICrop health detection with shareable PDF report
Two honest gaps. The AI crop-health feature assumed camera quality and connectivity it wouldn’t always have, and degraded ungracefully on 2G, an offline fallback like manual symptom selection should have shipped alongside it. And the buyer persona got far less attention than the farmer; with no PM to enforce persona equity, that discipline had to come from me, and it didn’t come enough.
I state plainly that the concept never shipped, so there is no post-launch data, inventing analytics for a project that never launched would be the easiest and worst thing I could do here.
What this project changed in me
KrishiConnect taught me that designing for low-literacy users takes away every shortcut I did not even realise I was relying on. There is no falling back on “they will figure it out from the label”, so every screen has to show its function before a single word is read. That constraint did not only help rural farmers. It made me a clearer designer for everyone.
In short
KrishiConnect is a self-initiated mobile concept that removes the middleman from Indian agricultural commerce, where farmers lose 15–20% of their crop value to information asymmetry. Designing for low digital literacy, 2G connectivity, and regional languages, I rendered live prices as trend graphs instead of numbers, replaced one overwhelming form with a 5-step conversational onboarding, and built an AI crop-health check with a shareable PDF. Aadhar-verified badges carry the trust, and the seven core features stay learnable. Proxy testing validated the graph-first bet, which is the kind of insight desk research would never have given me.