Frequently asked questions
Methodology, coverage, pilots. If something here doesn't answer your question, see how we measure or our notes.
Methodology
- How should I read your yield estimates?
- Our yield model is statistical — it trains on each county's history (past yields, current-season satellite NDVI, weather, soil context) and outputs an estimate with an expected error band. Read it directionally first: above-trend, on-trend, or below-trend. The point estimate carries known uncertainty that we report alongside it, and the band shrinks as the season progresses through silking and pod-fill. See the methodology page for the current model's validation results.
- How well does the model perform?
- Mid-season corn and soy yield modeling is genuinely hard at this lead time — the best academic models report substantial but not perfect skill. Our current 2025 corn.v1 beta model achieves R² 0.86 / RMSE 15 bu/ac at silking in 80/20 backtest. We treat directional accuracy (above/below trend) as the primary success metric over precise point forecasts. Numbers are refreshed each season; see /blog/reading-mid-season-yield-forecasts for the full breakdown.
- How do you pick the "neighbors" you compare my field to?
- For each county, we compute the great-circle distance from its centroid to every other county centroid and keep the 10 closest. Centroids come from the Census 2024 Gazetteer. A field on the IA/IL border can have neighbors in both states — the comparison crosses state lines when geography says it should.
Coverage
- How often does the data refresh?
- Sentinel-2 satellite imagery is captured every 5 days per location. We aggregate, mask to crop pixels, and publish weekly — one number per county per week. Production yield model predictions emit at six anchor weeks: W22, W26, W30 (silking for corn), W34 (pod-fill for soy), W38, W42.
- Which crops do you cover?
- Corn and soy in the contiguous US, anchored to USDA NASS county yield reports. We do not currently support pasture, specialty crops, wheat, or non-CONUS regions.
Pilots
- Is there a pilot or trial?
- Yes — we run 8–12 week pilots for producers and producer-hedgers who want to test mid-season directional reads on a portion of their operation. Reach out via the contact link on the pricing page to discuss scope.
- Can I export the data via API or CSV?
- No. YieldVista is a UI-first product — analytics stay in the dashboard. We don't offer CSV export, bulk download, or a public data API at any tier. The intelligence is the product; the data is just the input.
Limits
- Can I trade on this?
- YieldVista is observational data — not investment, trading, agronomic, or risk-management advice. We surface reads; how you use them is on you. See our full disclaimer for the legal version.