No measurement is perfect. OCR separates errors into two types.Random errors cause readings to scatter unpredictably about the true value, some too high and some too low. They come from things such as human reaction time when using a stopwatch, readings that fluctuate, or judging a position by eye differently each time. Random errors cannot be removed, but their effect can be reduced by repeating readings and calculating a mean, or by plotting a graph and drawing a line of best fit.Systematic errors shift every reading by the same amount (or the same proportion) in the same direction. Repeating the readings does not reduce them, because every repeat has the same error. Causes include a badly calibrated instrument, always reading a scale from the same wrong angle (parallax), or a zero error.A zero error occurs when an instrument gives a non-zero reading when the true value is zero, for example an ammeter needle resting at 0.02 A with no current. It is corrected by subtracting the zero reading from every measurement.Worked example: A micrometer reads +0.02 mm with its jaws closed. It then reads 1.47 mm across a wire. Corrected diameter = 1.47 − 0.02 = 1.45 mm.On a graph, random errors scatter points either side of the line of best fit. A systematic error such as a zero error moves the whole line, so a line that should pass through the origin has a non-zero y-intercept, although its gradient may be unchanged.