In short
Most weighing-site problems are not faults but operating habits: recording before the reading settles, a stored tare with no expiry, freely typed material and customer names, one shared login, correcting by deleting, a gate isolated from weighing, ignoring return to zero, and no tested backup. Each takes minutes to fix.
Most problems at weighing sites are not faults. They are operating habits that look acceptable day to day, and then show up at month end: reports you cannot rely on, differences with no traceable source, and disputes with customers that nothing can settle.
The following are repeated to the point of tedium, and every one of them takes minutes to fix — if you notice it.
Recording before the reading settles
The operator is rushed and the queue is long, so they record while the figure is still moving. The stored value can be hundreds of kilos out, and nothing on the ticket will ever say so.
The fix: do not allow recording until the reading is stable. If the system captures from the scale directly, make stability a condition rather than an option; if entry is manual, the written rule is simply: do not record until the figure has held for two seconds.
A stored tare with no expiry
A common shortcut: save the tare for regular vehicles and reuse it. The problem is that tare genuinely changes — fuel, equipment, mud, repairs — and after months the stored figure is fiction.
The fix: give stored tares an expiry, and flag any that differ sharply from the last real reading. Where possible, take an actual empty weight for each regular vehicle periodically.
Material and customer names typed freely
This is the single mistake that ruins reports without anyone noticing. “Sand”, “Washed sand” and “Washed sand ” with a trailing space become three different materials in the report. Six months later the materials report is meaningless and stock cannot be calculated at all.
The fix: a registered list of materials and customers that the operator selects from rather than types. Adding a new entry should need an admin, not happen mid-rush.
One login shared by everyone
It looks practical: one account on the machine and everyone works on it. But it removes the most valuable thing in the record — who did what. After any error or suspicion you cannot tell who recorded or who edited, and you cannot compare shifts.
The fix: an account per operator. This is not about trust; the record simply loses its value without it.
Correcting by deleting instead of adding
A ticket has an error, so it is deleted and rewritten. The real history is gone, and any later audit sees a clean version that does not reflect what happened.
The fix: record a correction as an addition — old value, new value, who changed it, when and why. Mistakes happen and must be correctable, but they should not disappear.
The gate isolated from weighing
Security logs arrivals and departures in a notebook, weighing lives in another system, and nobody compares them. So a simple question has no answer: which vehicle entered and was never weighed?
The fix: connect the two, and make exit conditional on a permit linked to tickets. This closes the most serious gap of all, as set out in the article on stopping fraud.
Ignoring the return to zero
The scale does not quite return to zero after a vehicle leaves, the operator gets used to it, and it becomes normal. It is not normal: it is dirt, standing water, or the beginning of real drift.
The fix: a daily glance at zero, and a note of any offset. What follows is in the calibration article.
No backup
A spreadsheet on one machine, or a local system with no copies. One disk failure erases a year's record — and with it any ability to prove your side of a dispute.
The fix: a daily automatic copy kept somewhere other than that machine. And more importantly: restore it once — an untested backup is not a backup.
The night-shift operator was never trained
Training happens once at installation. Then a new hire or a relief shift arrives and learns by imitation. That is where improvisation starts: different names, skipped steps, personal workarounds.
The fix: one page pinned in the control room with the steps in order, and a short induction for every new operator. It costs an hour and pays for a year.
The quickest wins
If you only start with three: a closed material list, an account per operator, and a tested backup. Together they take less than a day, and they change the value of every number you hold afterwards.