A person pipetting knows which tube is which, because they are looking right at it. A robot running hundreds of samples has no such intuition. If the link between a physical sample and its identity breaks anywhere in the run, the instrument will happily process the wrong specimen with complete confidence. Everything that keeps automation trustworthy at scale comes back to keeping identity firmly attached to every sample, from the moment it lands on the deck to the result at the end.
Barcodes carry the identity
A barcode, linear or a two-dimensional data matrix, is the usual anchor. It can sit on a tube, a rack, a plate, or all three, and it ties a physical object to a record in the LIMS. Tube-level barcodes identify individual samples, while rack and plate barcodes identify the container and, with a known layout, the position of every well inside it. Two-dimensional codes matter for small tubes and high density, where there is no room for a long linear barcode. The identity is only as good as the read, so unreadable or duplicate barcodes are a real failure mode, not an edge case.
Autoloaders and reading the deck
Many platforms read barcodes automatically as labware is loaded. An autoloader scans racks as they enter, confirming that what the method expects is actually present and where it should be, and refusing to run if a rack is missing or in the wrong place. Onboard scanners can read plate and tube codes on the deck itself. This is the step that catches the classic mistake, a rack loaded in the wrong position, before the robot acts on it rather than after. A run that begins with a clean scan of every container starts from known identity rather than assumed identity.
Positional tracking carries identity through the run
Barcodes establish identity at the edges, and positional tracking maintains it in between. Once the system knows sample X is in well A1 of a scanned plate, it follows that sample through every transfer: A1 to a new plate's B2, B2 into a pool, and so on. The instrument records the chain, so at the end you can say where any sample went and what was done to it. This is the difference between knowing what you loaded and knowing what happened to it, and it is what lets a result point back to a specific specimen through many steps.
It is really a chain-of-custody problem
Underneath the hardware, this is provenance. In a regulated or clinical setting the record of which sample was where, when, and what touched it is not a convenience but a requirement, and even in research it is what makes a surprising result investigable rather than mysterious. Identity tracking, the method that ran, and the liquid classes that moved each sample are all part of the same story: a defensible account of how a tube of sample became a row in your data.
The robot only knows what you told it. Scan identity in at the edges, track it through every transfer, and a result can always be traced back to the tube it came from.