What gets assessed is the task, not the farm
The question isn't “is my farm safe?”, but “what hazards is someone exposed to who carries out this particular task?”. Two farms with identical equipment can therefore end up with different results, because work is done differently on each.
That also explains why a template assessment pulled from the internet rarely holds up. It describes an imagined task, not your actual one. It's useful as a checklist of things to think about – taken over as a finished result, it describes a farm that doesn't exist.
The practical consequence is a breakdown. You split the work into tasks or work areas that make sense to consider together – milking, yard traffic, handling plant protection products, working at height – and assess each one on its own. A single document for the whole farm ends up either too coarse or unreadable.
Every task considered has two halves: the hazard identified and the measure derived from it. A hazard without a measure is an observation, not an assessment – and a measure without a named person responsible is a good intention.
It has to be kept up to date, not filed away
The assessment captures a snapshot. Once conditions change, it loses its validity – immediately, not just at the next scheduled review. A new machine, a remodelled barn, a changed work process, or new staff with different experience are all reasons to take another look.
Alongside that sits the regular review that happens without any particular trigger. It's the mechanism that stops an assessment from quietly going stale because a lot has changed in small steps, with no single step ever looking big enough to prompt action on its own.
In practice, keeping it up to date doesn't mean rewriting it from scratch. It means the assessment carries a version and a date, that it's clear who is responsible for it, and that a date exists on which someone will look at it again. Without that date, it's chance that decides when it's next opened – and experience says that tends to be after the accident.
The documentation is part of the duty, not just its proof
A common misconception is that you have to assess, and additionally document, so that you have something to show if it's ever in doubt. But the documentation is itself part of the duty – the result of the assessment, the measures decided on, and the outcome of checking them all have to be recorded.
The difference isn't academic. If documentation were only evidence, an undocumented but carefully carried out assessment would merely be hard to prove. As part of the duty itself, it's incomplete – even where the actual work on the farm is genuinely good.
For practice, that gives a simple yardstick: whatever isn't recorded well enough for another person to read and build on doesn't exist for the purposes of the duty. It's also why assessments that only exist in full inside the farm manager's head vanish at the next generational handover.
What sets recurring inspections apart from it
Alongside the risk assessment sit the inspections of work equipment: recurring technical checks on machines and installations that establish whether a piece of equipment is still in a safe condition. The two get confused easily, because both are about safety and both generate deadlines.
The difference lies in what's being looked at. The assessment looks at a task and asks what hazards arise from it. The inspection looks at a piece of equipment and asks whether it's still in the condition needed to be used safely.
The two are connected, though, and in one direction only: in Germany, it's the risk assessment that determines which inspections a piece of equipment needs and at what interval. Skipping the assessment therefore doesn't just leave out one duty – it also removes the basis on which inspection intervals are set at all.
And the everyday visual check before work starts is a third thing again: not a recurring inspection in the legal sense, but the kind of measure that can follow from the assessment. It replaces neither of the other two – but it does keep an eye on the gap between two inspection dates.