Anything that causes a measurement to deviate from its true value is called an error. We make a distinction between two types of errors: random and systematic.
Random errors are, well, random. They are thus unpredictable, inconsistent and difficult to reproduce. Because random errors keep changing in sign and magnitude, they show up as fluctuations in readings, and scatter in graphs.
Examples of random errors include
Since random errors are equally likely to be positive and negative, repeating measurements and taking the average value is the standard procedure to manage random errors. During averaging, chances are good that some positive errors will cancel out some negative errors. So the averaged value is likely to be close to the true value.
Systematic errors are, ahem, systematic. They are therefore predictable, consistent, and reproducible. Since systematic errors are fixed in both sign and magnitude, they do not show up as fluctuations in readings nor scatter in graphs.
Example of systematic errors include
Systematic errors cause a consistent deviation from the true values. So repeating measurements and averaging the readings does not help overcome systematic errors. On graphs, they do not show up as scatter. Instead, they may cause the best-fit-line to shift, or changes the trend of the data. To overcome systematic errors, we must first identify the cause of the systematic error, and then take appropriate corrective action to eliminate it. For example, after realizing that there is a zero error in your micrometer screw gauge, you can eliminate the error simply by recalibrating your MSG properly, or make appropriate adjustments to your readings.