When an assay looks noisy, the instinct is to blame the instrument. In our experience the instrument is rarely the largest source. Variance usually enters earlier, in the parts of the method that nobody writes down because they seem obvious.
Start with sample handling, not statistics
How a sample is sub-sampled, thawed, diluted and left standing on a bench can outweigh the precision of the detector. We log the interval between preparation and measurement for every run, because a two-hour gap is a different experiment from a twenty-minute one.
Write the acceptance criteria before the first run
If you decide after the data arrives what counts as acceptable, you have not defined a method — you have described a preference. Criteria set in advance also make it obvious when a run should be repeated rather than reinterpreted.
Repeat one control, not everything
A single well-placed control repeated across the plate tells you more about drift than triplicating every sample. We use one reference well per ten samples and plot it across the run; the shape of that line usually explains the outliers.
Report variance, do not hide it
Reporting the standard deviation of the method qualification is not an admission of weakness. It is the number your reader needs in order to know whether a small difference between two batches means anything at all.
In practice
- Log preparation-to-measurement intervals for every run
- Fix acceptance criteria in the study design
- Use a drift control rather than blanket triplicates
- Publish method variance alongside results
Noise is not a property of the instrument. It is a property of the whole procedure, including the parts you did not write down.
Illustrative article written for this template.
Tell us the question. We will scope the study.
Send a short description of your material, method and deadline. You will get a written response from a scientist, not a sales sequence.