Know what is in the tube: sample labelling, identifiers and chain of custody

A laboratory result is a statement about a particular sample. The measurement can be flawless, the instrument freshly calibrated and the method fully validated, and the result is still worthless if it belongs to a different tube than the one everybody thinks it does. Identification is the quiet foundation under every other kind of quality. It is also the part that tends to be improvised: a marker pen, a strip of tape, an abbreviation that made sense on the day, a box that was "always on the second shelf".

Reviews of laboratory error have pointed in the same direction for decades. Most mistakes do not happen inside the instrument. They happen before analysis begins, in collecting, labelling, transporting, splitting and storing material, and identification errors are a large share of them. They are also among the hardest to detect afterwards, because a mislabelled sample produces a perfectly ordinary looking number.

This guide covers how a laboratory makes sure it knows what it is holding. It walks through what a label needs to carry, the materials that keep a label readable through freezing, thawing and solvent wipes, how linear and two-dimensional barcodes work and fail, how a chain of custody record is kept, how derived samples inherit their identity, and how storage locations are mapped and audited. It ends with what to do when an identity is in doubt, and a checklist. It is written as a reference. No supplier, brand or service is named or recommended.

1. Why identity fails, and why it is so hard to see

Identification fails in a small number of recurring ways. A label is applied to the wrong tube because two were open on the bench at the same time. A label is written correctly but becomes unreadable, because the ink ran in a water bath or rubbed off under a gloved thumb. A tube is correct but its cap is swapped with a neighbour's, and the cap was the only thing labelled. A transcription error turns 0817 into 0871 when an identifier is copied by hand from one record to another. A box is moved to make room in a freezer, and the location record is never updated.

None of these failures announces itself. A smudged label at least looks wrong. A swapped label looks exactly as right as a correct one. That is why good identification practice is built on redundancy and on independent confirmation rather than on care alone. Care is necessary, but a system that only works when nobody is ever tired or interrupted is not a system.

A wrong label does not look wrong. The only defence is a second, independent way of knowing what you are holding.

2. What a label has to carry

The heart of any label is a unique identifier: a value that is assigned to one sample and to no other, ever. It should be generated by a system rather than invented on the bench, so that two people working on the same day cannot both create "Sample 12". Good identifiers are boring. They are sequential or random, of fixed length, and they carry no meaning that might later become untrue. An identifier that encodes a project name, a storage temperature or a person's initials breaks the moment the project is renamed, the material is moved, or the person leaves.

Around that identifier, a label carries just enough context for a person to notice when something does not fit. In practice that usually means a short description, the date of preparation, who prepared it, the storage condition, and for anything derived from another sample, its relationship to the parent. The full record lives in a notebook or a database. The label exists so that the tube and the record can always be joined, and so that a person can sanity-check the join at a glance.

Many quality frameworks ask for two identifiers at every point where a sample changes hands or is used. The idea is simple: one value can be wrong by accident, but two independent values are very unlikely to be wrong in a way that agrees. The second identifier can be a description, a date, a location, or a check digit, as long as it is checked rather than just printed.

The fields above are a common working set. The exact list depends on the laboratory's own procedures and on any standard it works to.

One rule deserves its own line: dates in year-month-day order. 03/04/26 is the 3rd of April in much of the world, the 4th of March in the rest, and in a shared freezer nobody knows which convention the author used. 2026-04-03 cannot be misread. It also sorts correctly, which matters the first time somebody exports a list of samples and orders it by date.

3. Labels that survive the conditions

Most unreadable labels were not applied carelessly. They were the wrong material for what happened next. A label in a laboratory may be frozen to very low temperatures, thawed with condensation on its surface, dropped in a water bath, wiped with alcohol, spun in a centrifuge and handled with wet gloves, all in one afternoon. Paper and ordinary office adhesives are not made for any of it.

Laboratory label stock is usually a synthetic face material such as polyester or polypropylene, printed by thermal transfer, where a heated ribbon fuses ink onto the surface. Direct thermal printing, the kind used on shop receipts, darkens heat-sensitive paper instead. It is fast and needs no ribbon, but the image fades with heat and light and can be wiped off by some solvents, which makes it a poor choice for anything kept longer than a few days.

The adhesive matters as much as the face. An adhesive rated for deep cold is formulated to stay tacky at temperatures where ordinary adhesive turns brittle and lets go. Almost all of them share one requirement: apply the label at room temperature to a clean, dry surface, and let it set before the tube goes into the cold. A label pressed onto a frosted tube straight from the freezer is bonded to a film of ice, and it falls off when the ice melts.

Ratings differ between manufacturers. The only reliable test is to label a few blank tubes and put them through the laboratory's real worst day before relying on the stock.

Handwriting still has its place, for a single tube or a temporary transfer. When it is used, it should be in a marker made for laboratory use, on the body of the tube rather than the cap, with the full identifier rather than a nickname. Caps are interchangeable. Tubes are not. A label that exists only on a cap is a label waiting to be swapped.

4. Barcodes: linear, two-dimensional, and verified

A barcode is not more accurate than a person. It is more consistent. A person reading a twelve-character identifier will eventually transpose two digits. A scanner reading a well printed code essentially never will, and it reads in a fraction of a second. That is the whole case for barcodes: they take the most error-prone step, copying a value by eye, out of the process.

Linear barcodes, the familiar row of bars, are common on the side of tubes. In laboratories the usual symbology is Code 128, which encodes letters and digits compactly and includes a built-in check character, so a scanner can tell when a read is corrupted. A linear code needs a clear margin on both ends, called the quiet zone, and it needs to be printed so the bars run along the length of the tube on narrow containers. Wrapped around a small diameter, the outer bars curve away from the scanner and the code stops reading.

Two-dimensional codes, most often Data Matrix in laboratory work, pack the same information into a small square. They tolerate damage well, because the data is stored with error correction and the code can be read even with part of it scratched or frosted. Their biggest use is on the base of tubes held in racks, where a flatbed reader images an entire rack at once and returns the identifier in every position. That turns a slow, error-prone task, writing down what is where, into a single scan.

Whatever the symbology, two habits separate a barcode system that works from one that only looks modern. The first is printing the human-readable value from the same data as the code, never typing it separately, so the two cannot disagree. The second is scanning the label back after it is applied. A barcode that has never been scanned is an assumption. Reading it once, on the tube, before the tube goes anywhere, proves that the printer, the label, the placement and the record all agree.

5. Chain of custody

A chain of custody is the documented history of who has held a sample, from the moment it is created or received until it is used up or disposed of. Every transfer is a link: one person or location hands the sample to another, and the record says when, between whom, and in what condition it arrived. When every link is present, anyone can follow the sample's path and see that nobody unaccounted for ever held it. When one is missing, that stretch of the sample's life is unknown.

A useful custody record answers the same handful of questions at every link. Who released the sample, and who accepted it. The date and time of the transfer. The identifier, checked against the paperwork rather than copied from it. The condition on arrival, including temperature where it matters and whether any seal was intact. Both parties sign, or both are recorded by a system that knows who is logged in. A transfer that only one person recorded is a claim, not a link.

Tamper-evident seals extend the chain across stretches where nobody is watching, such as a courier journey or a weekend in a shared freezer. A numbered seal is applied over the closure or the outer bag, its number is written on the custody form, and the receiver checks both that the seal is intact and that the number matches. A seal that has been broken and replaced, even with good intentions, should be recorded as exactly that.

Formal chain of custody is mandatory in some fields, such as forensic and regulated environmental testing, and optional in many others. Even where it is not required, a light version pays for itself the first time somebody asks where a sample has been. The minimum is a log of transfers with dates and names. The rest can be added when the stakes require it.

6. Aliquots and derived samples: parents and children

Many samples do not stay whole. They are split into smaller portions, aliquots, so that each portion can be thawed once and used without disturbing the rest. They are diluted, extracted, filtered and combined. Each of these steps creates new material that needs an identity of its own, and the identity has to keep its link to the material it came from.

The cleanest approach is to issue every derived portion its own unique identifier and record the parent in the database, rather than building lineage into the identifier itself. A scheme such as S-1234-A, S-1234-B looks tidy until somebody splits S-1234-B again and the suffixes start to nest. A separate parent field never runs out of room, and it allows a full family tree to be drawn: which original sample, which intermediate, which final portion, and which results came from each.

Whatever the scheme, two rules hold. First, label the child before the material moves. A portion that sits unlabelled in a fresh tube while the next one is being prepared is exactly the moment two tubes get confused. Second, record the split as one event: one parent, a known number of children, the volume or amount in each, and the time. If the record says six portions were made and the rack holds five, someone needs to know before the day ends, not months later.

7. Storage maps and inventory

A sample is only useful if it can be found. In most laboratories that means a location hierarchy: a building and room, a freezer or cabinet, a shelf or rack, a box, and a position inside the box. Each level gets its own identifier, ideally a barcode, so that a location can be recorded by scanning rather than by memory. "Freezer 3, rack C, box 12, position F7" is a findable sample. "Probably in the old freezer" is not.

Boxes with a fixed grid, each position addressed by row letter and column number, make the record precise. They also make a missing sample visible: an empty position that the record says should be full is a discrepancy that can be investigated. Racks of tubes with two-dimensional codes on the base go a step further, since a rack scan reconciles every position against the record in seconds.

Records drift from reality unless they are checked. A periodic inventory audit, sometimes a full count and more often a rotating sample of boxes, compares what the record says with what is physically there. Every discrepancy is logged and resolved, never quietly corrected. A record that is edited to match the freezer without explanation hides the very error the audit was meant to find. Removals should be recorded as they happen, and samples that are used up or discarded should be closed in the record, not left as phantom entries that send the next person on a search.

Storage units also fail. A freezer that alarms at night may need to be emptied into another unit in a hurry. Laboratories that cope well with this have planned for it: spare capacity, a written procedure, and a quick way to record where every box went. The same location hierarchy that makes everyday retrieval easy makes an emergency move recoverable.

8. When an identity is in doubt

Sooner or later somebody will pick up a tube and not be sure what it is. The label may be damaged, two tubes may look identical, or the location record may disagree with the label. The response should be the same every time, and it should be written down before it is needed.

• Stop and separate. Put the tube aside, physically apart from anything it might be confused with, and mark it as quarantined.
• Do not relabel from memory. A confident guess written onto a new label turns an uncertain identity into a false certain one.
• Gather the evidence. Check the location record, the preparation log, the custody record, neighbouring positions, and any partial label remains.
• Decide on the record. If the identity can be established from independent evidence, document how. If it cannot, the sample is treated as unidentified, and its fate is decided by the procedure, not by convenience.
• Find the cause. One unreadable label is an accident. Several from the same batch or printer is a process problem worth fixing.

It is tempting to resolve doubt in favour of keeping a sample that took weeks to produce. That is exactly why the decision belongs to a written procedure. A sample of unknown identity can still look valuable, but any result produced from it carries the same uncertainty forward into every conclusion it touches.

9. A practical checklist

The list below is a starting point for reviewing how a laboratory identifies and tracks what it holds. It is general, and any real procedure should be adapted to the laboratory's own work, the standards it follows and the conditions its samples go through.

• Identifiers are generated by a system, unique, of fixed length and free of meaning that can expire.
• Every label shows the identifier both as a barcode and as text, printed from the same data.
• Dates are written year-month-day, everywhere.
• Labels go on the tube body, never only on the cap.
• Label stock and print method have been tested against the real worst-case conditions.
• Labels are applied at room temperature, to a dry surface, before freezing.
• Each label is scanned back once after it is applied.
• Two independent identifiers are checked at every transfer and before every use.
• Every hand-off is recorded with time, sender, receiver and condition on arrival.
• Derived portions get their own identifiers and a recorded parent.
• Every storage level, from unit to position, has its own identifier.
• Inventory is audited on a schedule, and discrepancies are logged rather than silently corrected.
• There is a written procedure for samples whose identity is in doubt.

None of this is glamorous, and none of it shows up in a result. That is the point. When identification works, nobody notices it. When it fails, every measurement downstream inherits the failure, and nothing in the numbers will say so.