Putting a soil science tool in the hands of the people who need it

In short

A state agency had a tool that produced excellent soil health reports, but only people who code could run it. I built the front door.

The situation

The Washington State Department of Agriculture had already solved the hard part. Jadey Ryan, a data scientist there, had built an R package called soils that turns soil sampling data into a report a farmer can act on. Programs in other states wanted it for their own farmers.

The catch was who could run it. The package needed R, a local install, and a command line. The farm advisers who sit across the table from farmers have none of that. So every report went through someone who codes.

What I did

Jadey and I had known each other through the R-Ladies community for years, and in 2024 she asked for help with the last layer. We built Dirt Data Reports, a web app that sits on top of the package. An adviser downloads a template spreadsheet, in English or Spanish, uploads their data, fills in a few fields about their project, and presses a button. The app checks the data, hands it to the package, and returns the report as a web page or a Word file.

The important decision was what not to build. The science stays in the package Jadey maintains. The app only turns clicks into the function calls a data scientist would have typed, so when the package improves, the app improves with it.

It went live in September 2025. Programs in California, Idaho, and Oregon now run reports for their own farmers. And the places where advisers pushed past what the app expected became the clearest list anyone had of what the package needed next. Jadey and I told the story at posit::conf 2026.

What changed

Most organizations have a tool like this. It works, and only one kind of person can use it. The fix is rarely a new tool. It’s the layer that meets the next person where they are.