“Information alone rarely changes behavior.”
The insight behind Nudge (Thaler & Sunstein, 2008)
The control room for farm advisory
See who acted. Reach who didn't.
Outturn is where advisory partners — NGOs, agri-input companies, government programs — run their farmer programs: send weather-timed WhatsApp nudges, watch follow-through by district, and re-nudge the rest.
Not a passive dashboard — the room where you see and act.
Latur cotton cohort · this month
28 of 42 farmers acted
What it does
Outturn runs on a multi-modal, closed-loop advisory engine — it reaches farmers by text and voice on WhatsApp. The layer we built is what makes every piece of that advice accountable: tracked, rolled up, audited, and re-sent when it slips.
Every WhatsApp reply resolves to Done or not — per farmer, per cohort.
DynamoDB Streams aggregate every reply into per-district follow-through summaries, in real time.
A Next.js dashboard where partners watch follow-through and re-nudge who slipped — keyless to AWS via OIDC.
Every cohort, license and reminder cycle recorded — not editable after the fact.
How it works
A nudge is one timely WhatsApp message telling a farmer to act now, in their own language — like the short dry, low-wind window when spraying actually works. Outturn watches each district's weather, sends only when that window opens, and follows up until the farmer confirms or it closes. Tap a step or watch it play.
The differentiator
Table stakes · every advisor does this
The closed loop · only Outturn
Peers like Farmer.Chat, iSDA, and AgriChat deliver the advice. None we found documents the closed loop.
For partners
A partner is any organisation that advises farmers at scale — an NGO, an agri-input company (seed, fertiliser, crop-protection), a co-operative or FPO, or a government extension body — a KVK (Krishi Vigyan Kendra, the farm-science centres that carry advice to India's villages). Each gets its own branded, licensed workspace; the farmers just use plain WhatsApp.
How a partner runs it
Group a district’s farmers by crop and language — say an NGO’s 26 cotton growers in Latur. The engine starts watching that district’s weather.
Switch the cohort live on a Starter, Growth, or Scale plan. Checkout runs through Stripe.
See confirmed action, not delivery receipts — Latur at 67%, 28 of 42 nudges acted on.
Re-nudge the farmers who haven’t acted — the 14 in Latur who didn’t — by hand or on a schedule.
Every plan is the full closed loop: WhatsApp advisory, audited follow-through, and the monitoring dashboard. You choose the tier by how many farmers you reach, and the cost per farmer falls as you grow. Each cohort activates through Stripe checkout.
up to 50 farmers
₹999 / month
≈ ₹240 / farmer / year
up to 250 farmers
₹2,999 / month
≈ ₹144 / farmer / year
up to 1,000 farmers
₹9,999 / month
≈ ₹120 / farmer / year
Why it matters
Good crop advice already exists — it just rarely arrives in time. With far too few extension workers to reach everyone, no one is there at the moment it counts. And in farming, the moment is everything: spraying only works in a short dry, low-wind window. Miss it and the pass is wasted — the pest spreads, and the crop is lost.
So the bottleneck was never knowing what to do. It's getting it done — the right action, at the right moment, across thousands of farmers at once.
The scale of the gap
~126M
smallholder and marginal farmers in India, about 86% of all holdings.
Agriculture Census 2015-16Outturn closes that gap — advice tracked all the way to done, so a missed window becomes a re-nudge instead of a lost crop.
How it's built
Three planes, one system. The AWS-and-WhatsApp engine sends nudges and captures replies, a single Amazon DynamoDB table records and rolls up every outcome, and the Vercel control plane reads those roll-ups and re-nudges the engine over keyless OIDC. Not a dashboard bolted onto a database.
The engine, and the green layer H0 built on it
Green is what H0 added; grey is the AgriNexus engine that pre-dated it — a deep, all-serverless AWS stack (9 Lambdas, Step Functions, Bedrock, the lot) that won the AWS AIdeas Innovation Award. H0 owns two green planes: the Vercel control plane (multi-tenant dashboard, Stripe licensing, re-nudge) and the Amazon DynamoDB data layer — the single-table schema, Streams→Aggregator with DEDUPE# idempotency, the SUMMARY# roll-ups and a separate Marketplace-provisioned audit table. Vercel reads it all over keyless OIDC.
Built on
The engine and data
The control plane
The farmer's channel
Want to look under the hood? See the full-resolution proof — the live AWS & Vercel consoles, the single Amazon DynamoDB table, and every screen behind the loop, at full quality.
What's next
See it live