Category: QC Basic Concepts | Estimated reading time: 8 minutes
Building Intuition, Not Just Definitions
The concept that there are two types of variation is likely already familiar. But understanding the definitions alone is not enough—what is more important is building the intuition that you will use every time you face a QC reject in the laboratory.
Because at the bench, no one explicitly asks, "is this a random or systematic error?" But a trained medical laboratory scientist (MLS) will read the graph patterns, observe the distribution of points, and intuitively know where to look. That intuition is what we aim to build here.
Two Names, Two Worlds
Westgard (2016) defines both quite strictly.
Random Error (RE) is an error component that varies unpredictably from one measurement to the next. Its direction cannot be predicted—sometimes high, sometimes low, with no consistent pattern. In statistics, random error is reflected in the standard deviation (SD) and expressed as CV (Coefficient of Variation) in percentage. The larger the CV, the less precise our measurement system is.
Systematic Error (SE) is an error component that is consistent and unidirectional—always higher than the true value, or always lower. It repeats with the same pattern from run to run. In statistics, systematic error is reflected in the bias, expressed as d% (percentage deviation from the reference value). The larger the d%, the less accurate our measurement system is.
Its relationship with Total Error becomes clear here:
TE = |d%| + 2CV
Random error contributes through CV. Systematic error contributes through d%. Together, they form Total Error—a measure of the overall uncertainty of every result we release.
The Easiest Analogy to Understand: Weighing Yourself
Imagine you weigh yourself every morning. Your actual body weight is stable at 60 kg, unchanged for a week.
Random Error Scenario: Monday: 60 kg. Tuesday: 62 kg. Wednesday: 59 kg. Thursday: 61 kg. Friday: 58 kg.
The numbers fluctuate every day even though your weight hasn't changed. There is no consistent pattern—sometimes higher, sometimes lower. Your scale is not precise. The causes could be many small things: slightly different standing positions, an uneven floor, or a slightly unstable sensor.
Systematic Error Scenario: Monday: 62 kg. Tuesday: 62 kg. Wednesday: 62 kg. Thursday: 62 kg. Friday: 62 kg.
The numbers are consistent—but always 2 kg higher than the actual weight. There is something systematic causing this bias. Perhaps the scale was not calibrated correctly from the start, or there is a hidden load on the platform.
The most fundamental difference: random error makes you confused because the numbers are unstable. Systematic error makes you wrong because the numbers are stable—but consistently wrong.
The latter, in many ways, is more dangerous. Because it looks convincing.
Direct Implications in the Laboratory
The scale analogy represents exactly what happens in the laboratory.
When CV is high (random error dominates): if you measure the same control material 10 times and get values that jump far from the mean, your CV is high. This indicates inconsistency in the process—it could be from imprecise pipetting, insufficient sample homogenization, incubation temperature fluctuations, or operator variation. Investigation: check all operational conditions that can change from measurement to measurement.
When d% is high (systematic error dominates): if all 10 measurements are consistent—but consistently above or below the target value—your d% is high. Something is systematically causing bias: drift in calibration, a new reagent lot that has not been verified, degraded control material, or an issue with the measurement method itself. Investigation: check factors that are permanent or semi-permanent—calibration, reagents, storage conditions.
Ricós et al. (1999), in the biological variation database that serves as an international reference, emphasize that separating random and systematic components in laboratory performance evaluation is the foundation of rational QC planning.
On the Levey-Jennings Chart, the Differences Are Clear
The Levey-Jennings chart is the most powerful visual tool to distinguish between these two types of errors directly.
Random error will appear as points jumping irregularly around the mean. There is no consistent direction. Statistically, this is the expected appearance of a normal distribution—as long as the amplitude is not too large.
Systematic error will appear as a trend (points moving up or down gradually) or a shift (points suddenly moving to a new level and staying there). Both patterns are signals that something has changed systematically in the process.
Westgard, Barry, Hunt, and Groth (1981), in the classic paper that introduced the multirule system, explicitly designed QC rules based on this difference: some rules are designed to detect random error (1₃s, R₄s), while other rules are specifically designed to detect systematic error (2₂s, 4₁s, 10x). We cannot use all rules indiscriminately without understanding what type of error we are looking for.
Why This Is Important for Daily Practice
For MLS professionals who have worked for many years, this concept may already be intuitively familiar—even if it hasn't been labeled with formal terms.
Have you ever experienced a situation: QC rejects, you repeat it, the result gets back in—but the next day it rejects again with the same pattern? That is most likely a systematic error. Repeating the measurement will not solve this problem, because the root cause lies in the calibration or the reagents, not in the individual measurement.
Or conversely: QC rejects once, you repeat it and it passes immediately, and it doesn't happen again anytime soon. That is more consistent with a random error—likely due to a conditional factor that occurred only at that moment.
Recognizing these patterns is not just a technical skill. It is a diagnostic ability that distinguishes a reactive MLS from a proactive MLS.
The distinction between random and systematic error is also closely related to the concept of measurement uncertainty—Braga & Panteghini (2020) emphasize that a proper understanding of bias (systematic) and imprecision (random) components is essential for correctly estimating measurement uncertainty in modern laboratories.
Conclusion
Two types of errors, two ways to read them, two ways to investigate them. Not meant to be memorized—but to be internalized until they become instinct.
Because in a busy bench, there is no time to open a book. All you have is the ability to read patterns and make the right decision in minutes. That ability starts with a solid understanding of what is actually happening within our QC data.