Two things get measured, and they give very different answers. Both are real; the difference between them is the honest picture.
| What was measured |
Sales |
Typical miss |
Within 20% |
| You type a parcel number and let GAIA find everything itself — the normal way the tool is used | 61 | 33% | 31% |
| GAIA is handed a verified soil rating, sale-date acreage and township | ~200 | 24% | — |
Read it this way. Most of GAIA's error is not the pricing model — it is getting the inputs right. When the soil rating and acreage are correct, GAIA is roughly twice as accurate. That is also why a number you have checked the acreage and soil on is worth much more than one you have not.
Against our appraiser. When GAIA is given the same soil rating, acreage and tillable split our appraiser used, it lands within 10% of the appraiser's value on about 3 of 4 row-crop farms. From a parcel number alone, with nothing checked, it lands within 10% on about 2 of 5 row-crop farms and about 1 in 4 farms of any kind, and misses by about 18% on average. So treat GAIA's figure as a range, not a single number: most of all for woods and recreation land, tracts under 20 acres, appraisals of part of a parcel, counties with no appraisal history, and any parcel where the acreage and soil have not been checked.
For a lender. GAIA is a screening and scoping aid, not a desk-review substitute and not a bankable figure. Roughly three in ten valuations land within 20% of the eventual sale price on the everyday path, so treat any single number as a starting point for review. Farmland itself carries about 20% irreducible price scatter between equally reasonable buyers, so no model of any kind gets far below that.
Measured 2026-08-31 against independent recorded sale prices, driving the live product end to end. Method and full results: GAIA validation dossier, phase 2.
