Suppose {y_t: t = 1, 2, …, n} is a time series and n terms of the series are y_t, y₂, …, y_n respectively. Taking the time t on X-axis and the term y_t of the time series on Y-axis, the point (1, y₁), (2, y₂) … (n, y_n) are plotted on the graph paper. Then points are joined by line segments. The graph so obtained is called the graph of the time series. A continuous curve passing from the viscinity of most of the points is drawn. The curve so obtained is called the trend line of the time series.

This is the simple and crude method of determining the trend of the time series. This method is quite easy to understand. But the mathematical form of the trend cannot be obtained by this method. When the plotted points of the time series are widely scatter from one another, it is difficult to draw an unique curve representing the trend of the time series. In such a situation, more than one curve can be drawn representing the trend of the time series. As a result it becomes difficult to determine the trend of the time series.