Farm records have a particular failure mode. The information exists — it is in a diary in a bakkie, a WhatsApp message to the mechanic, a feed invoice in a drawer and somebody’s memory of which cow was treated. None of it is wrong. It is just not anywhere you can use it when you need a number.
It is easy to conclude that farms keep poor records because nobody writes things down. In practice the opposite is true — people write a great deal down. The problem is that recording happens in the field, where there is mud, gloves and frequently no signal, while the system that wants the data lives in an office.
Any approach that depends on someone remembering four hours later, at a desk, will lose data. It always does.
Service intervals, repairs, fuel, documents and suppliers. The number most farms cannot produce on demand is the true operating cost of a specific machine — purchase, depreciation, fuel, parts and downtime combined. Without it, “should we repair or replace” is a feeling rather than a decision.
Individual animal identity, lineage, breeding and births, health events, treatments and medicine records, feed and weights. Medicine records deserve particular care: withdrawal periods have direct food-safety and market consequences, and “I think it was about three weeks ago” is not a record.
Every stage from land preparation to yield, with the costs attached. Yield alone is vanity. Yield against input cost per hectare is the number that changes what you plant next season.
Populations, breeding, permits, health and sales. Permits especially — these carry hard deadlines and the penalty for missing one is not a gentle reminder.
Time, seasonal teams, approvals and documents. Seasonal labour is where informal record keeping most often becomes a compliance problem.
Feed, parts, stock levels and stocktakes. Knowing you have three drums of something is useful. Knowing you will run out on Thursday is what prevents a standstill.
1. It has to work with no signal. Coverage on a farm is patchy by definition. If recording a treatment requires a connection, it will not be recorded. The record must save on the phone and sync later, without anyone thinking about it.
2. Identification should be scanned, not typed. An ear tag, a sticker on a drum, a plate on a tractor — scanning the code and having the right record open removes both the typing and the transcription error. This matters more than it sounds: most bad farm data is a mistyped identifier, not a missing entry.
3. Workers need a different view from managers. A worker does not need the margin analysis. They need today’s jobs, a way to record what they did, and a way to report a problem — behind a PIN, in plain language. Give a general-purpose management system to a field worker and you get either no data or bad data.
Do not digitise everything at once. Pick the record whose absence costs you most — for many farms that is equipment servicing or livestock medicine — and get that one complete and reliable for a season. A single trustworthy record beats six half-maintained ones.
FarmBase is built around exactly these constraints: equipment, livestock, crops, wildlife, workforce and stock in one system, tag scanning to pull up the right record, a PIN-protected worker mode for the field, and everything continuing to work when the signal drops.