Writing a Standard Operating Procedure: A Template That Survives Being Followed

A standard operating procedure exists so that a competent person who was not present when the work was designed can perform it and reach the same result. That is the whole test, and the only one worth applying to a draft. Small laboratories usually write procedures because somebody asked for them, and what results is a descriptive essay about how the work is generally done — accurate, readable, impossible to follow. This page describes the other kind: a document written as instruction, and versioned so that anyone can establish which text was current on a given date.

What an SOP is for, and what it is not

A procedure does three jobs. It makes an operation repeatable, so two runs a month apart differ because of the material rather than the operator. It makes it transferable, so competence does not leave the building when a person does. And it makes it defensible, so a result can be shown to have come from a stated method rather than from whatever happened that afternoon. Procedures held under document control are how testing laboratories demonstrate that their methods are stable over time 1.

Be equally clear about what an SOP is not. It is not a training document: it assumes a competent operator and does not teach pipetting or how a balance works. It is not a rationale document; reasons interleaved with instructions make the instructions harder to follow. And it is not a wish. A procedure describing what the laboratory would like to do is worse than none: it documents a gap between stated and actual practice, which is precisely the finding an assessor looks for.

Scale matters here. Almost every available template assumes a regulated manufacturing site with a quality unit and staff whose entire job is document control. A laboratory of three people has none of that and needs none of it. What it needs is a short document, under version control, saying what is actually done. The requirement is that a procedure be identified, current, followed and retrievable 4, not that it pass through four signatures.

The five sections that carry the document

Five sections do the work. They are named differently in different houses but answer the same five questions, and content fitting none of them is usually background or policy rather than procedure.

Scope and records are the sections most often treated as formalities, and the two that decide whether the document is usable. A scope stating who may perform the procedure settles in one line a question otherwise answered inconsistently. A records section written alongside the procedure guarantees that every step producing a decision also produces evidence of it, which is what makes a run reconstructable long afterwards 3.

Writing steps somebody else can execute

The procedure section is read while wearing gloves and should be written for that. Seven conventions carry most of the benefit.

• One instruction per step, one verb per instruction. If a step contains the words and then, it is two steps.
• Number every step, and never renumber on revision without a change record. People cite step numbers in notebooks and in investigations; silent renumbering makes every such reference wrong.
• Use the imperative, addressed to the operator. Record the balance reading, not the balance reading is recorded — the passive voice conceals who is responsible.
• Put the condition before the action so the reader can skip the step: if the solution is cloudy, filter it.
• State quantities with units and tolerances every time. Ten millilitres plus or minus half a millilitre is a specification; about ten millilitres is a hope.
• Name decision points explicitly and say what to do at each branch, including the branch where the answer is no.
• Say where each value is written down, in the step that generates it.

The commonest defect is not error but vagueness, and it is systematic: the author knows what is meant, so the ambiguity is invisible to them alone. The test that catches it is having another person perform the procedure from the text, with the author present but silent. Every question asked is a defect, fixed by editing the document rather than by answering the question.

Version control, effective dates and review

Version control answers one question: which text was current when this work was done? Everything else is administration. Give each procedure a short permanent document number and a version incrementing on every issue. Put both in the header block and both into the record the procedure produces, so a completed record points unambiguously at the text which governed it.

Distinguish the issue date, when the document was written, from the effective date, when it must be followed. A week between them gives operators time to read a change before it binds them. Record what changed and why in a short revision line rather than in tracked changes. A revision history reading minor updates is worthless six months later; one reading step 7 filtration interval changed from 10 to 20 minutes following incomplete filtration on two runs explains itself indefinitely.

Withdraw superseded versions from the point of use rather than leaving them beside the current one. An obsolete copy in circulation is the commonest way a laboratory ends up following a procedure it has already replaced. Retain the superseded text in an archive; reconstructing old work needs it. The principle governs any record carrying evidential weight: the original is preserved, and its supersession documented rather than concealed 2.

Set a review interval — two years suits most laboratory procedures, annually for anything stability-critical — and review on trigger as well. Triggers are where real revisions come from: a failed run, a new instrument, a changed supplier, and above all the discovery that everybody has quietly stopped following a step. That last case demands a decision. Either the step matters and practice returns to it, or it does not and the procedure loses it. Leaving in a step nobody performs converts the whole procedure into fiction, and an operator who has learned to ignore one step will ignore the next.

Worked skeleton: goods-in receiving

What follows is a complete skeleton for the procedure most laboratories need first and write last: receiving material into storage with its identity intact. Copy it, change the numbers, delete what does not apply.

Scope. Applies to all research material received, from unpacking through to storage and creation of the batch record. Excludes consumables and bulk solvents. May be performed by any trained laboratory member; the batch record entry is checked by a second person before first use.

Materials and equipment. Delivery documentation; the certificate of analysis; the batch register; label stock rated for the intended storage temperature and a solvent-resistant marker; the storage unit temperature log; the quarantine shelf.

1. Record the date and time of receipt and the name of the person receiving the delivery.
2. Inspect the outer packaging before opening. Record its condition and whether cold chain packaging was still cold. If packaging is damaged or coolant fully thawed, quarantine and raise an investigation before proceeding.
3. Open the packaging. Retain every document supplied with it.
4. Record the batch or lot reference from the container label.
5. Compare the label against the delivery documentation and the certificate of analysis. Record the comparison as performed. Any mismatch is recorded as such and quarantines the material.
6. Transcribe from the certificate: identity, molecular weight and the form it refers to, net peptide content, purity with its method conditions, counterion, date of analysis.
7. For any required value the certificate omits, record the omission, then any figure assumed in its place and why. Never leave an assumed value indistinguishable from a stated one.
8. Assign the batch its internal identifier and mark every container with it.
9. Place the material into storage within 30 minutes of opening. Record location to slot level and condition as applied.
10. Complete the register entry. Mark the material available for use only after a second person has checked it against the certificate.
11. File the certificate against the batch record rather than in correspondence, with its analysis date alongside.

Acceptance criteria. Material is accepted when label, delivery documentation and certificate agree on identity and batch; the certificate states purity with its method conditions; storage was reached within 30 minutes of opening; and the register entry has been second-checked. Any criterion not met quarantines the material, recorded with its reason.

Records. One register entry per batch carrying every field transcribed in steps 4 to 8; the retained certificate; the container labels; and any investigation record. The entry carries the document number and version of this procedure. All entries are made at the time the step is performed, in ink, by the person performing it.

That skeleton runs to about a page and a half, and it is the shape every other procedure in a small laboratory can take. The parts that look bureaucratic — the second-person check, the recorded comparison, the flagging of assumed values — are the parts that pay, because each converts a silent failure into a visible one at the moment it occurs rather than months later.