Rainfall cover,
priced by a market

Assessing a two-hectare farm costs more than the policy is worth, so smallholders cannot buy drought cover. Mmiri removes the assessor: five underwriter agents bid against each other for the risk, and the rainfall record settles it.

Rainfall in this window, 1981–2025

Sokoto (13.06, 5.24) — 30 days from 15 June
0200400mm2006 · 47.381.6 mm strike198519901995200020052010201520202025
Below the strike — the policy paysAt or above — no payout
See the 45 figures behind this chart
Window total by season — Sokoto (13.06, 5.24), 30 days from 15 June
SeasonTotalOutcome
1981139.2 mmNo payout
198280.7 mmPaid
1983130.2 mmNo payout
198467.5 mmPaid
1985134.6 mmNo payout
1986107.7 mmNo payout
198760.5 mmPaid
1988193.2 mmNo payout
1989117.4 mmNo payout
199098.1 mmNo payout
1991126.8 mmNo payout
199281.6 mmNo payout
1993142.8 mmNo payout
1994154.6 mmNo payout
199571.8 mmPaid
1996118.0 mmNo payout
1997143.6 mmNo payout
1998101.0 mmNo payout
199971.4 mmPaid
2000100.3 mmNo payout
200192.0 mmNo payout
200279.4 mmPaid
2003150.4 mmNo payout
200469.2 mmPaid
2005229.1 mmNo payout
200647.3 mmPaid
2007112.6 mmNo payout
2008147.8 mmNo payout
2009101.0 mmNo payout
2010119.5 mmNo payout
2011109.7 mmNo payout
2012153.6 mmNo payout
2013103.4 mmNo payout
201448.4 mmPaid
2015104.9 mmNo payout
2016210.4 mmNo payout
2017225.0 mmNo payout
2018126.2 mmNo payout
201997.9 mmNo payout
2020114.9 mmNo payout
2021176.8 mmNo payout
2022152.1 mmNo payout
2023154.9 mmNo payout
2024442.7 mmNo payout
2025257.9 mmNo payout

9 of 45 seasons since 1981 fell below 81.6 mm and would have paid. 1 of those 9 happened since 2011.

Seasons that would have paidSokoto, every season on record since 1981
9of 45
Driest window in the record2006 — against a 81.6 mm strike
47.3mm
Underwriters competing for the riskeach with its own capital, its own key, and its own read
5
Policies settled on testnetone of them by a TEE oracle reading NASA directly
3

How one round works

A sealed-bid auction, start to finish, in about two minutes on chain.

You ask for cover

One place, one 30-day window, your maximum price — escrowed with the request.

Agents bid blind

Each underwriter submits a hash of its premium. No one can read a rival's bid.

Bids are opened

After the window closes, each reveals. The contract rechecks every hash.

The lowest price wins

Cheapest premium under your maximum takes the risk, and locks the full payout.

Rainfall decides

The oracle reads NASA. Below the strike the payout moves — no claim is filed.

4°N6°N8°N10°N12°N14°NEnuguNsukkaIbadanMakurdiKadunaKanoSokotoMaiduguri

Markers sit at the coordinates the oracle sends to NASA. The country outline is a schematic. Pick one to reprice the page.

Eight places, eight different risks

Rainfall is local. Enugu and Maiduguri sit in different climates on the same map, and their windows do not even start in the same month — the south plants in April, the far north in June. A single national price would be wrong everywhere.

The market prices each site on its own record, so the premium follows the place rather than the country. Pick one on the map and every figure on this page recomputes for it.

Set your terms, then buy the cover

Change anything on the left and every figure recomputes from the committed rainfall record. The price on the right is what the market would charge you now.

30 days from 15 June
Default is the 20th percentile of this site’s prior seasons
The winner locks this in full as collateral

What the five agents would quote

50 USDC payout — 81.6 mm strike — 30 days from 15 June 2026
AgentIts view of the historyBreach rateLoadPremium% of payout
uw-2 Recent would winOnly the last 15 seasons tell you anything; the old ones are a different climate.11.8%20%7.0614.1%
uw-5 CapacityBlend the views, then ration the last of the capital by charging more for it.16.5%20%9.9119.8%
uw-3 BlendSplit the difference and charge a little more for not knowing which is right.16.5%25%10.3320.7%
uw-1 LongrunEvery season since 1981 counts equally; the climate is noise around a mean.21.3%20%12.7725.5%
uw-4 CautiousPrice the worse of the two views. Survive first, compete second.21.3%35%14.3628.7%
At these terms the market clears at 7.06 USDC — 14.1% of the payout — written by uw-2 Recent. The spread between cheapest and dearest is 7.30 USDC, which is the price of the disagreement below.

Why this needs a market and not a formula

9 of the 45 seasons on record at Sokoto ended like the ground behind this text.

20.0%All seasons since 1981fell below 81.6 mm in 9 of 45 seasons
6.7%Last 15 seasonsfell below 81.6 mm in 1 of 15 seasons
0.33×The gaphow much more often it has happened lately than across the whole record

Both numbers are correct. They disagree, and the premium you should charge differs accordingly. No single formula can know which rate is right, because the question — how much does the distant past tell you about next season — is a real open question about a changing climate. So Mmiri does not pick. It runs an auction, and the underwriters that misprice the risk lose their own capital.

Settled on testnet, not described

Every one of these is a transaction you can open and a reading you can recompute yourself with pnpm verify-oracle.

Policy 7

Paid out
53.6 mm

below the 60.0 mm strike — 50 USDC went to the holder. Settled via the keeper path.

Policy 9

Paid out
53.6 mm

settled by an OutLayer worker inside a TEE, not a trusted account. Settled via the OutLayer path.

Policy 8

Expired
113.7 mm

well above the strike — no payout, collateral released. Settled via the keeper path.

Connected to market.mmiri-demo.testnet. Opening a round means a NEP-141 ft_transfer_call carrying your maximum premium — see rounds.
Rainfall is NASA POWER PRECTOTCORR (Precipitation Corrected), daily, community=AG. Totals are computed in integers and rounded once, so every implementation agrees. One grid cell is roughly 56 × 69 km near Enugu — rain on the cell is not rain on your farm, and that basis risk is inherent to index insurance.