Replicate Variance Check Tool — Free Mean, SD & CV Checker

This tool checks replicate agreement: for each set of repeated measurements, it reports the mean, the sample standard deviation (SD), and the coefficient of variation (CV% = SD ÷ mean × 100), flagging sets whose CV% exceeds 20% as high variance. Replicates exist to answer one question — would the same measurement, repeated, give the same answer? — and CV% answers it in scale-free form: a CV of 2% means tight agreement whatever the units, while a CV of 25% means the replicates disagree enough to distrust the mean. You paste replicate sets and the tool flags the problem rows.

How to use it

Paste one replicate set per line, values comma-separated.
Click Check replicates.
Read mean, SD, and CV% per set.
Investigate every high-variance flag before using that mean.
Re-measure or exclude flagged sets with a documented reason.

Worked example

Replicates 10.1, 10.3, 10.2 give mean 10.2, SD 0.10, CV 1.0% — tight agreement, use the mean with confidence. Replicates 5.0, 8.0, 6.0 give mean 6.33, SD 1.53, CV 24.1% — flagged high, and rightly so: the spread is nearly a quarter of the mean, so the mean hides disagreement rather than summarizing it. The lab’s next step is not to average harder but to ask why: pipetting error, a heterogeneous sample, an unstable assay — the flag starts an investigation, and the investigation, not the average, fixes the data.

Honest limits

The 20% CV threshold is a common screening rule, not a universal law: acceptable CV depends on the assay, the concentration range, and the field — some methods run happily at 30%, others must stay under 5%. CV also misbehaves near zero means (dividing by a tiny mean inflates CV) and with very few replicates, where SD itself is unstable. The flag is a prompt to look, not a verdict to delete: never drop a replicate set on a flag alone without a documented technical reason.

Who it is for

This tool is for lab workers checking assay replicates, students learning precision versus accuracy, and researchers screening data before analysis. It is a quality-control flag, not a data-cleaning service.

Going further

Track CV% over time, not just per run: a rising CV trend across weeks signals a degrading instrument, a tired reagent, or drift in technique, long before any single run crosses 20%. Plot the CV of a control sample on a simple run chart, and the week the points start climbing is the week to service the instrument — before data quality collapses. Replicate checks feel like overhead when every run passes; they pay for themselves the first time a flag catches a bad run before it enters a dataset that would have taken weeks to unpick. Do not average replicates before checking them. The mean of a tight set and the mean of a wild set look identical in a table, and once averaged, the disagreement is gone from the record unless the CV was reported alongside. Make the check a step in the workflow: compute CV first, flag, investigate, and only then average the sets that passed, reporting the CV for every reported mean. Readers who see a mean with its CV can judge precision for themselves; readers who see a mean alone must trust you blindly. In fields where replicate counts are small, also report the individual values in a supplement, because three numbers carry more information than any summary of three numbers ever will. When a set is flagged, check the raw instrument output before blaming technique: a single mistyped digit, a misplaced decimal, or a swapped sample label creates exactly the pattern of one wild replicate among tight ones, and transcription errors are both common and cheap to fix.

Embed this tool on your website — free

Copy the code below to add this working tool to your own page. A small credit link stays with it — that’s all we ask.