The site description and the land mix, before you open the file.

One report, four logins

Everything in the panel above is something you already assemble by hand before the analysis starts. The difference is where it comes from:

Four sources, four calls, one response shape. No screenshots, no exports, and no step where a number gets retyped into a document you sign and certify.

The land mix is the analysis

Cropland at 100%, CRP at 90%, non-cropland that is taxed but generates no revenue at 10%, public right-of-way at 0%. That allocation is what turns a sale price into a $/tillable CSR2 point — the only unit on which farmland sales are actually comparable, because productivity is normalized by the unit itself. Get the allocation right and only market conditions, size, and location remain to adjust.

The problem is not the method. It is that appraisers are not consistent in how they break land mix down, which makes shared sales data unreliable the moment two people work the same file. One method, applied the same way to the subject and to every sale, is worth more than a better method applied unevenly.

• CDL classes allocated to cropland, CRP/grass, and non-cropland, reconciled to deeded acres less right-of-way
• Deeded, right-of-way, effective (taxable), and weighted effective acres reported separately instead of collapsed into one number
• Cropland CSR2 computed by excluding the map units you don't farm, not estimated off the whole-tract rating
• The same derivation on the subject and on every comparable, so the grid is internally consistent

Note the Poweshiek subject above. Four verified sales bracket it, but the 80-acre auction tract adjusts well above the median on a $/point basis — a small-tract premium, not a market signal. The workbench flags the spread instead of averaging through it, because reconciliation weighs the quality of the data and an unexplained average is not a conclusion.

Same boundary, same method, same table — on the subject and on every sale in the grid.

A workfile that answers the question

“The soil survey” is not provenance. Under the Record Keeping Rule the workfile has to support the report, and a reviewer asking where a cropland CSR2 came from wants a source, a retrieval date, and a derivation that still reproduces in five years — ten, if the file goes to litigation.

Because every figure in the panel above is computed from the subject geometry rather than captured from a screenshot, the workfile entry is the call itself: the layer, the parameters, the response, and the date. Re-run the same boundary and you get the same table. That is a materially different position to be in during a review than a folder of cropped PDFs.

Same subject, same boundary, same answer, three years from now.

Built to sit under your stack

We are not writing your report and we are not signing it. If you have a form platform or a narrative template your clients accept, an adjustment grid you defend in review, and a scope of work you have refined over years, keep all of it. Same for comps — if you maintain your own sales file, LandHero is another verified Iowa source to score against rather than a replacement.

REST, JSON, one schema across every county we cover. Most shops are pulling live parcel data the same week — and the first thing they automate is the site description nobody enjoys writing.