You can randomize a small trial by hand, and many have. Doing it well means separating the random sequence from the people who enrol participants. This guide shows the manual methods, spells out how they fail, and draws the line where software becomes the safer choice.
Free sandbox · No credit card · 21 CFR Part 11 aligned
| Sequence is unpredictable | Allocation is concealed | Audit trail | |
|---|---|---|---|
| Coin toss or dice at the bedside | |||
| Alternating or by day of the week | |||
| Pre-generated list held by the enrolling clinician | |||
| Independent person, opaque sealed numbered envelopes | |||
| Independent central service or validated system |
Randomizing without software in brief
The purpose
Randomization allocates participants to treatment groups by chance, so that known and unknown prognostic factors tend to be balanced between arms, and so that the people involved cannot choose who gets what. If a clinician steers sicker patients into the arm they believe is better, the comparison is biased before the first dose, and no analysis can fully repair it.
Good randomization has two parts, and each can fail on its own. The first is generating an unpredictable sequence. The second is concealing it from the person who enrols participants until the participant is irreversibly in the trial. A perfectly random list that sits open on the coordinator's desk is not concealed. Reporting standards such as CONSORT ask trials to say who generated the sequence, what type of randomization was used and any restrictions such as blocking or stratification, what mechanism was used to conceal allocation and who had access to it. Check the current CONSORT statement for the wording of the items.
So the question for a manual method is not only how the list was made. It is who holds it, when it is opened and how you could show afterwards that the rules were followed.
A manual method
This is the usual route for a small two-arm trial. It is not a recommendation for any particular study; your statistician and your sponsor decide.
Fix the number of arms, the allocation ratio (1:1 is common) and whether you need to balance on a factor such as site or disease severity. Write it into the protocol and the statistical analysis plan. Changing the design after the list exists is a way to introduce bias.
Simple randomization (a coin flip per participant) can produce badly unbalanced groups in a small trial. Permuted blocks keep arms close to equal: for a block of four with two arms, each block has two of each assignment in random order. Use varying block sizes, such as 4 and 6, and do not tell enrolling staff the block size, because a fixed block of four lets anyone who knows it predict the last assignments.
Someone who will not enrol participants or assess outcomes, ideally the trial statistician or an independent colleague, creates the sequence using a reproducible method such as a published random number table or statistical software with a recorded seed. File the seed, date and method in a document kept apart from the study team. The randomization list generator on this site can help you check a design and block structure; it is a planning tool, not your concealed study list.
The classic method is sequentially numbered, opaque, sealed envelopes prepared by the independent person. Each envelope holds one assignment, is numbered in sequence and is opened only after the participant has been enrolled and has consented. Lining the envelope with foil or carbon paper, and sealing across the flap, reduces the chance of someone reading the assignment through the paper. Alternatively a pharmacist or an off-site colleague can hold the list and give the assignment by phone or message when enrolment is confirmed.
Specify when an envelope may be opened (after eligibility is confirmed and consent is signed), who opens it, what to do if one is opened by mistake and how to log the date, time, participant ID and envelope number. Envelopes are used strictly in order, and unused envelopes are accounted for at the end.
Log each allocation in the study record at the time it is made, with date, time and the person who made it. A paper log is acceptable for a small study, but its integrity relies on people, not on a system.
Compare the list with the allocation log and the data. Check that envelopes were used in order, that the arm counts match what you expect from the blocks and that no assignment changed. Report any deviation and how it was handled.
The honest part
These are well-known failure modes. Whether any occurred in your trial is hard to prove after the fact, which is part of the problem.
| Failure | How it happens | Effect | Mitigation |
|---|---|---|---|
| Predictable sequence | Alternation, fixed small blocks, or a known pattern | Enrolling staff can guess the next assignment and steer participants | Random permuted blocks of varying size, never disclosed |
| Broken concealment | Transparent envelopes, envelopes opened early, list left in the clinic | Selection bias and an inflated treatment effect | Independent holder; opaque numbered envelopes; open only after consent |
| Out-of-order use | Envelopes skipped or discarded after peeking | Sequence disturbed without a record | Strict order; account for every envelope |
| Transcription and arithmetic errors | Wrong arm recorded or list mis-copied | Participants receive the wrong treatment | Second check; log at the time; reconcile at the end |
| No stratification | Balance on site or severity cannot be guaranteed by hand | Imbalance on key prognostic factors | Separate lists per stratum, or use software |
| Weak audit trail | Paper logs can be changed or lost | No proof for inspectors or reviewers | Move to a system with an audit trail for regulated work |
Capture includes randomization in the same system as EDC, eConsent and ePRO, with one audit trail. Try it with sample participants in the free sandbox, no credit card needed.
The line
Manual randomization is defensible when all of the following hold: the trial is small, low risk and run at one site; there are two arms with simple allocation; no one is blinded in a way that depends on concealed allocation; and nothing is being submitted to a regulator as pivotal evidence. Many pilot and feasibility studies sit here. Even so, the checklist above has to be followed carefully.
Software becomes the better choice when any of these appear. Multiple sites, because envelopes have to be physically managed at each one and central control disappears. Blinded designs, because the system can keep treatment labels from blinded roles. Stratification or minimization, which need many interlocking lists or a calculation after every participant. Drug supply links, where allocation must match kit numbers. Regulatory submission, where an inspector expects a record of who allocated whom and when. Large enrolment pace, where the chance of human error rises. For comparison of options, see clinical trial randomization software, block randomization software and randomization software for small studies.
Software does not remove the need for design decisions. Someone still has to choose the method, set the strata and decide who is unblinded. What it adds is enforcement and a record: the next assignment is released only when the rules are met, and the event is logged with a time and a user.
In Capture
Capture includes randomization alongside EDC, eConsent, ePRO and safety reporting, in one system with one audit trail. That means consent, allocation and the first case report forms sit in the same record, and teams do not need a separate randomization vendor or a spreadsheet for a simple design. For the specifics of your design, such as stratification factors or the allocation ratio, test them in the sandbox or ask for a demo before you rely on them.
Day 28 is an Arm B only visit, so Arm A participants never see it.
Reporting it
Whatever method you use, describe it so a reader can judge the risk of bias. Say who generated the sequence and how, what type of randomization and any restriction such as blocks of varied size were used, how allocation was concealed, who enrolled participants and who assigned them. If envelopes were used, say that they were opaque, sealed and sequentially numbered. If something went wrong, such as an opened envelope, say so and say what you did. The CONSORT flow diagram generator helps lay out participant flow, and the sample size calculator helps you size the trial before randomization starts. If you have no independent holder available, see how teams run randomization without a CRO.
Yes, for small, low-risk, single-site trials. A usual manual method is permuted blocks of varying size generated by an independent person and concealed in opaque, sealed, sequentially numbered envelopes opened only after enrolment. It carries risks of predictability, broken concealment and weak records.
Generate a random sequence with a random number table or statistical software, keep it with someone independent of enrolment, and release each assignment only after the participant is enrolled. Use permuted blocks if you need balanced arm sizes in a small trial.
A coin flip produces a random assignment, but it provides no concealment or record, and the person flipping often knows the result before the participant is committed. It is generally not considered adequate for a trial that others will rely on.
Allocation concealment means the person enrolling a participant cannot know or predict the next assignment until the participant is in the trial. It is separate from blinding, which concerns whether people know the assignment after allocation, and it is a separate item in reporting guidelines like CONSORT.
Envelopes can be held up to light, opened early, discarded or used out of order, and there may be no way to prove it did not happen. Opaque, sealed, sequentially numbered envelopes prepared by an independent person reduce but do not remove the risk.
Typically for multi-site trials, blinded designs, stratified or minimization allocation, linked drug supply, regulated submissions and fast enrolment. These need enforcement and an audit trail that a manual process cannot give.
Yes. Capture includes randomization in the same system as EDC, eConsent and ePRO, with one audit trail and blinded roles that never receive treatment-arm values. Confirm your exact design, such as strata and ratios, in the free sandbox or a demo.
A list generator creates the sequence, which solves only the first half. You still need a way to conceal the list and release assignments in order with a record. The generator on this site is a planning tool and is not your concealed study list.
Keep exploring
Randomization list generator
Check a block design.
Clinical trial randomization software
What software does.
Block randomization software
Blocks and balance.
Randomization for small studies
When size is the constraint.
Randomization without a CRO
Run it in-house.
EDC with randomization
Allocation in the data system.
Blinded and unblinded roles
Keep labels away from blinded staff.
Statistical analysis plan guide
Write the analysis plan next.
Free sandbox with every feature. No credit card, and you pay only when you go live with real participants.