Readability is not accuracy, and most Bangladeshi laboratories buy more of the first than their room can support. What the ceiling fan, the bench and the sample temperature are doing to your last digit, and a ten minute test that shows you.
A new balance is unpacked, set on the nearest free bench, switched on, and the last digit never settles. The usual response is a warranty claim. The usual cause is the room.
Readability is the size of the smallest increment the display shows. It is the number in the catalogue and it is the number people compare. It says nothing about whether the instrument can reproduce a reading.
The specifications that decide that are repeatability, usually given as a standard deviation, and linearity. A balance with a readability of a tenth of a milligram standing on a wooden bench under a ceiling fan will happily display a tenth of a milligram and repeat to several times that. The display is not lying. It is showing an increment it can resolve but cannot hold.
The useful consequence is a calculation almost nobody does. Take the repeatability you actually measure in your own room, decide the relative uncertainty your method needs, and from those two work out the smallest sample mass you are entitled to weigh on that instrument. That number, not the readability, is what should be written on a card taped to the balance. Weighing a sample smaller than that limit is the commonest way a laboratory produces a precise looking result that means nothing.
It follows that buying more readability than the room can support is wasted money. A soils laboratory weighing moisture tins needs a balance that settles quickly in a dusty room far more than it needs another decimal place. Match the instrument to the smallest thing you genuinely weigh, then spend the difference on the bench it stands on.
Vibration reaches a balance through the bench, and in Bangladesh it arrives from three predictable directions: the campus generator that starts every time the grid drops, footfall from a class arriving, and a compressor or centrifuge sharing the same benchtop.
The arrangement that works is unglamorous. Ground floor if possible, in a corner, against a load bearing wall rather than a partition, with the balance on a heavy top that is mechanically separate from the working bench around it. A stone or thick concrete slab on its own supports, not bolted to the bench frame, removes most of what a light laminate top passes straight through. Where a purpose built balance table is not available, a benchtop isolation platform does the same job: the Thorlabs passive benchtop isolation range covers pneumatic platforms, bolt-on isolator feet and sorbothane feet for exactly this class of moderately sensitive instrument.
Then check the bubble. A balance slid along a bench, or set on a floor that is not flat, is out of level, and that is a systematic error no calibration corrects. It takes five seconds and it is the step most reliably skipped.
Ceiling fans are universal in Bangladeshi laboratories and they are the commonest reason a reading will not settle. So is a split air conditioner blowing across the bench, a door onto a corridor, and an open window in a room cross ventilated in the cooler months.
The draught shield only works closed. A shield with a door left open, which is how most balances are actually used between weighings, is decoration. Teach the sequence: place, close, wait, read.
The subtler air problem is the sample itself. Anything warmer or cooler than the weighing chamber sets up a convection current around the pan and reads wrong until it equilibrates. A moisture tin straight from the drying oven is the classic case. Tins go into a desiccator and come out at room temperature, and the wait is part of the method, not a delay in it.
Humidity works the other way here. In a room at Dhaka's usual humidity a hygroscopic powder gains mass while it sits on the pan, and the reading climbs steadily with no apparent fault. Weigh those materials in a closed vessel and by difference rather than open on the pan, and record the room humidity alongside the result so that an odd figure can be explained later.
Before deciding a balance is faulty, run this. It takes about ten minutes and it will usually name the culprit.
The number the laboratory is entitled to report is the one that survives the worst of those conditions, because that is the condition the room is in when a student is actually weighing. If the spread with the fan on is unacceptable, the answer is not a better balance. It is a switch on that fan, a different corner of the room, or a heavier bench.