Probably a report you can name and cannot put your hand on. Between the soil test and what comes out of the fertilizer truck there is a chain of three links, recommendation, purchase and application, and a farm that never wrote the first link down cannot defend the other two, no matter how good the analysis it paid for.
Ask on any farm where this year’s blend came from and you get a short silence and then three answers: it came from the analysis, it came from what has always been done here, and it came from what the neighbor is running. All three can be true at the same time and nobody can separate them. The field that changed crops keeps getting the previous crop’s blend. The new field gets the blend from the field beside it, because they look alike. One field has been getting a heavier rate since a year the result came back poor, and that report is gone, so the rate stayed.
What it costs arrives the moment somebody outside asks. The buyer, the lender or the auditor wants to know what the farm bases its applications on, and the answer is that the farm tests its soil, which is true and is not an answer. The owner pays the same price from the inside: at the end of the cycle there is no way to tell whether what was spent had a document behind it or a habit.
What one written plan puts in order
A written plan holds that chain in one document and starts it one step further back, at the analysis, keeping the last link inside the plan instead of in a separate logbook. The Wisconsin version of conservation practice standard 590, published by the United States Department of Agriculture Natural Resources Conservation Service in September 2005 and written for that state alone, lists what a nutrient management plan has to contain. Three items sit in a row: the results of the soil, plant, manure or organic byproduct sample analysis, then “recommended nutrient application rates”, then “documentation of actual nutrient applications” carrying the rate, form, timing and method.
The line after the third item is the one worth copying onto a farm no standard reaches. “Revise the plan to reflect any changes in crops”, the standard says, naming yields, tillage, management and the soil or manure analyses as the other changes that trigger the rewrite. The same list asks the plan to carry guidance for implementation and for keeping records.
Reading that as law would be the wrong use of it, and outside Wisconsin it is not law at all. Read as evidence, it says something that survives every border: people who had to make this defensible in front of a third party decided that the analysis, the recommendation and the application belong in one document, in that order, and that the document gets rewritten whenever any of the three moves. Every standard in Production has that same second half, and it is the half that gets dropped.
Why the date on the report belongs in its own column
A recommendation is only as current as the analysis underneath it. The Voluntary Guidelines for Sustainable Soil Management, approved by the FAO Council in December 2016 and written for a broad audience rather than for one jurisdiction, ask that fertilizer application methods, types, rates and timing be based on soil and plant analyses, and that this be “a long-term endeavor rather than short term action”.
The second half of that sentence is what turns a date into a column. A single report used once is short-term action wearing the clothes of evidence, and it reads exactly like the other kind on the invoice. A report that is the current one in a series is what the guidelines describe. The only way to tell them apart on your own farm is the date written next to the document name, which is where the field’s soil test series stops being a formality.
The recommendation did not arrive on its own
The soil test is one input to the decision, and the people who study the decision keep finding the rest of them. O’Connell and Osmond interviewed 105 farmers in Missouri, North Carolina and Ohio between 2015 and 2016 and reported nutrient management decisions drawing on yield data, market prices, weather, product availability, prior experience, risk management, university research reports, soil and tissue sampling, and social networks. Soil sampling is one item on a list of nine. Those were producers in three states of one country, heard about a decade ago, so what carries over is the shape of the decision and not the proportions.
Rees and colleagues surveyed 2,728 Swiss arable farms across late 2023 and early 2024, and treated the same plurality as something to write down. Advice from an agricultural adviser, from an agricultural retailer, from a cantonal or national institution, from other farmers, from social media and from publications are six separate variables in that dataset, and the structural description of each farm records “whether the soil has been assessed and a soil management plan exists”.
That last item takes a second to read for what it is. Whether the soil has been assessed and a plan exists is a fact about a farm, recordable in a column, with a yes and a no. A survey reaching thousands of farms has to ask it because it cannot be assumed, and the farm that cannot answer it about itself is answering it anyway.
One basin already asks the question with a form
In one of these markets the question already comes with paperwork attached. Uruguay’s Ministerio de Ganadería, Agricultura y Pesca describes an action plan for water quality in the Río Santa Lucía basin whose third measure requires, of every rural holding inside that basin, control of the application of nutrients and fertilizing on the basis of soil analysis. That is Uruguayan law, it covers one basin, and it obliges nobody anywhere else.
What crosses the border is not the rule. It is the demonstration that the question can be answered with paper the farm already owns. Somebody drafted a requirement a farm satisfies by pointing at a report, which means pointing at a report is a thing a farm can be asked to do, in front of a regulator or in front of a buyer who has no regulator behind them.
The document you filled in yourself
The report is an answer, and the question was written by the farm. The soil analysis submission form used by the soils laboratory of Embrapa Amazônia Oriental, printed in Documentos 266 in 2006, exists to guide the laboratory in interpreting the data and in the liming and fertilizer recommendation. Owner, property, the sample number and the hectares it stands for, what grew there the year before, what is planned: the recommendation is built on top of what the farm declared before the sample left.
That makes the origin document locatable by construction, and it also names a gap nobody goes looking for. Where the farm kept no copy of the form it sent, the report can still be found and half its origin cannot. What the laboratory was told about that field is gone, and the recommendation reads as a number with no question above it, which is a different problem from having no report. The soil test report by field is where the report’s own fields get read.
Five columns, one row per field
The sheet is short by design, and only the fifth column produces work. Two of them are transcription: what was applied is already written on an invoice or a work order somewhere on the farm. The third and fourth are the ones never written down, and the fifth is a check in the sense that the four functions of management use the word, a comparison between a standard and a fact that ends in one of two words.
| Column | What goes in it | Where it comes from |
|---|---|---|
| Product | The blend or the material, as the invoice writes it | Purchase invoice, work order |
| Quantity | Rate and total, in the units the invoice uses | Purchase invoice, application record |
| Source document | Name and date of the report or of the signed recommendation | Report folder, adviser’s file |
| Who recommended | Name of the person or of the firm | The same document, or a phone call |
| Checks, or no document | One word | Whether you could put your hand on it |
A filled line reads like this: field 7, a 4-14-8 blend, 260 kilograms per hectare (232 pounds per acre), report for field 7 dated March 2025, recommended by the name signed on it, checks. The line under it reads: field 12, same blend, same rate, no document. Nothing on this sheet says whether either rate was a good one, and that is the point of it. The sheet asks where the decision came from, and a different competence entirely asks whether it was right.
Where to start, and what closing it looks like
An hour and a half, with the invoices for this cycle, the folder of reports, and the phone number of whoever made the recommendation. Nothing to buy and nothing to install.
The second sheet is the one that changes anything, and it will be shorter than expected. Most of those lines close with a phone call, because the person who made the recommendation still has the file when the farm does not. What that turns up is uncomfortable in its own way: the origin of a decision the farm executed and paid for is sitting on somebody else’s computer, and it stays there until somebody asks for a copy.