A model that lives with the orchard
Every season leaves evidence behind. The next decision should begin with what this block has already taught us, not with a blank forecast.
The future Harvora is building is closer to the ground: an intelligence layer that remembers each block, translates uncertainty into consequence and gets sharper every season.
A document for growers · 9 pages · Harvora Ltd
Harvora should feel less like opening a tool and more like continuing a conversation with the land. It should know what changed, what mattered and what the last season taught us, then make the next uncertain decision easier to see.
Each season adds another layer of context.
Every season leaves evidence behind. The next decision should begin with what this block has already taught us, not with a blank forecast.
A probability is only useful when its range, its reason and its consequence are visible. Harvora should be able to argue both ways in plain English.
Public data gets us to the block. The next layer brings the ground, canopy and microclimate into the same living picture.
Every season should leave the next one better informed.
A recommendation should arrive with its reasoning attached.
Cherries · Lapins · 3.2 ha · Tuesday night
Predicted minimum −1.4 °C. A 70% probability of breaching your damage threshold. 80% confidence interval: −0.8 °C to −2.1 °C.
Recommended: run wind machines from 11pm to 4am. Ask Harvora why, and it walks you through the cost-loss maths, which models agree, and how this block behaved the last time conditions looked like tonight.
The technology can be complex. The relationship with the grower cannot be.
A frost night is not a data problem. It is a grower awake at 2am, a crew rung before dawn, and a call that gets defended to a family and a bank in the morning. Harvora is built for that room: growers, crews and the land are all part of the brief. If a screen does not change what someone does before sunrise, it has no reason to exist.
Judged on the night, not the dashboard
Your block’s history is yours. We do not sell it, fold it into someone else’s product, or use it to price you. And a sharper decision should show up as restraint: a frost fan that never ran, a spray that was not needed, diesel still in the tank come morning.
Your data leaves with you if you leave
A 70% that is right seven times in ten beats a confident number that is right half the time. So the scoreboard stays in the open: every eligible outcome is checked against what actually happened, while unsupported predictions remain marked as awaiting validation.
Eligible predictions are scored; unsupported ones stay visible
Harvora does the arithmetic nobody has time for at 10pm: a thousand simulated seasons, each one priced. What it cannot know is that the southern corner always goes first, or that the crew is two people short this week. So it hands over the number and stops. The call belongs to whoever has to live with it.
Nothing acts on the orchard by itself
An answer that lands after the fans should have started is not late, it is worthless. Everything here is built backwards from the moment of the decision: on the phone while there is still time to act, not in a report that explains the loss the next morning.
In hand before the decision, or it does not count
Not a black box. Not another layer of noise. A living record of the land, the uncertainty and the call, becoming more useful every time the season asks a difficult question.
Weather, ground and canopy signals enter the record.
The block keeps the evidence from each season.
Uncertainty becomes a clear, priced call.