Advisers & research

Running on-farm trials and studies with growers (a researcher's guide)

FieldSafe team · Draft, scheduled

FieldSafe study map with photo clusters and an arrivals timeline
A study map with photo clusters from participating farms and an arrivals timeline.

An on-farm trial compares treatments on real farms, under real conditions, with the grower’s own team doing the work. In FieldSafe, a researcher sets up a field trial with 2–12 treatments plus a control and 1–24 visits by growth stage or date, invites farms by email, and watches visits and photos arrive on one map. Each farm chooses how precisely its location is shared, can leave at any time, and is told plainly that exported data cannot be recalled.

This guide walks through the parts that make or break a multi-farm trial, and where a tool helps.

Designing a fair on-farm comparison

FieldSafe records the trial. It does not design it for you, and it does not run trial statistics. The design is the researcher’s job, and it is worth getting right before the first farm is invited. A few general principles of good practice:

  • Always have a control. The control is what the farm would do anyway. Without it, you cannot tell a treatment effect from a good season.
  • Compare like with like. Put a treatment and its control on Blocks, or parts of Blocks, with the same cultivar, age, rootstock and irrigation where you can. Two Blocks that differ in more than the treatment will tell you about the difference, not the treatment.
  • Repeat across farms. One farm is a case study. Several farms let you see whether an effect holds across soils, slopes and managers.
  • Decide what you will measure, and when, before you start. Write the protocol: which observations, at which growth stages, with which photos. Changing it mid-season makes the data harder to compare.
  • Keep it workable for the farm. A grower who has to remember a complicated rotation of plots will make mistakes. Fewer, clearer treatments usually give better data than many muddled ones.

For deeper design advice, such as how to assign treatments to plots and how many replicates you need, work with a statistician or your institution’s trial guidelines. Treat FieldSafe as the collection and reporting side of the work.

Setting up the study

In FieldSafe, a researcher chooses between an observation study (watching and recording, with no treatments) and a field trial. A field trial has:

  • 2–12 treatments plus a control;
  • 1–24 visits, each set by growth stage or by date;
  • protocol files attached, so every farm works from the same document;
  • a one-page setup.

Visits by growth stage

Calendar dates are easy to plan but don’t suit every crop. Flowering on a warm, low-lying avocado farm near Hazyview can be weeks ahead of a cooler farm near Haenertsburg. If you ask both to record “flowering” on 15 August, one of them is recording the wrong thing.

That is why each visit can be set by growth stage or by date. A growth-stage visit says what to look for, and the farm records it when its trees get there. The farm plans the visits itself, fitting them around its own work.

To keep things moving, FieldSafe sends automatic visit reminders to farms, at most one every 3 days, so a farm is nudged without being nagged.

Offline data capture

Trial visits happen in the orchard, and orchards often have no signal. In FieldSafe, a visit is recorded on a phone like any other field work: the observations, plus detail photos. It is saved on the phone and sends itself when there is signal. Nobody has to walk back to the office to upload.

On the researcher’s side, the study workspace shows:

  • Home, with every participating farm on one map and what came in today;
  • the study map, with photos placed at each farm’s chosen precision, clustered by tag, visit, treatment, date, farm or crop stage, and an arrivals timeline you can replay;
  • Photos in a lightbox, Compare to set treatments side by side, Patterns and Farms;
  • private photo notes and labels, which only the study team sees.
FieldSafe compare treatments side by side in a study
Comparing treatments side by side in a study, using the photos and visits farms recorded.

You can invite co-coordinators to share the work, such as a colleague or a postgraduate student who labels photos.

Trial data comes from someone’s farm. Growers are more willing to take part, and to keep taking part, when the rules are clear from the start.

How farms join

Farms are invited by email. The link lasts 30 days, and the farm can preview the study before signing in. A farm joins through a short join page, without having to set up the whole of FieldSafe.

If you change the study, for example adding a visit, it becomes a new version. Each farm then accepts the new version or leaves. Nobody is moved into a study they did not agree to.

What the farm controls

The farm choosesOptions
Location precisionExact, Block centre, or About 1 km
When visits happenThe farm plans its own visits
Whether to stayIt can leave at any time

What the data never carries

  • No person names in exports. Researchers never see who took a photo.
  • The PDF report never draws a coordinate.
  • Push notifications never show names or notes.

The honest part

Data that was exported cannot be recalled. If a farm shares visits and the researcher downloads them, that copy exists. FieldSafe tells farms this plainly before they share, and researchers should say it too. It is better for a grower to hear it at the start than to discover it later.

For questions about consent forms, ethics approval or how data may be used in publications, follow your institution’s ethics process and agree the terms with each farm in writing. FieldSafe gives you the controls; it does not replace that agreement.

Reporting back to farms

At the end of the season, the researcher downloads:

  • CSV or JSON data (54 columns), with a data dictionary that explains each column;
  • a ZIP of photos with checksums, so you can confirm no file was changed or lost;
  • a GeoJSON map file;
  • a PDF report with treatment-versus-control charts.
FieldSafe study report and data package export
The study report and data package export: data, photos with checksums, a map file and a PDF report.

The PDF is the natural thing to send back to growers. A short summary of what the trial looked at, with the charts and a few photos, is often the reason a farm says yes to next season’s trial. Note that the charts show what was recorded; any statistical analysis is done by the research team in their own tools.

A practical example

Nomsa runs a fictional biostimulant trial on avocado for a research group, across nine farms between Tzaneen and Hazyview. She sets up three treatments plus a control, with four visits by growth stage: flowering, fruit set, fruit at golf-ball size, and pre-harvest.

She invites the farms by email. Seven choose “Block centre” precision and two choose “About 1 km”. At each farm, the Supervisor records the visit on a phone in the orchard, with detail photos of the fruitlets, and it sends when the bakkie gets back into signal.

Nomsa watches arrivals on the study map. When she adds a fifth visit in February, it becomes a new version: eight farms accept it, and one leaves. At season end she downloads the data package and the PDF report, analyses the data with her statistician, and sends each farm the PDF with a one-page summary of what she found.

Invite your farms

FieldSafe is in early access and is tested on a staging environment with a few test users. If you run trials or observation studies across several farms, apply for early access and tell us about your next season. Read more on the researchers page, how photos are kept on photos, what an avocado farm records on the avocado page, and how data is protected on security.

FieldSafe helps you prepare; certification is decided by your certification body.

Early access

Be one of the first farms on FieldSafe.

FieldSafe is in early access. Join, and we help you set up your farm and try it on your team's own phones.

Join early access