How the coefficient is computed
The Pearson coefficient is measured on log returns, not on prices. Correlating prices produces huge and wrong numbers: two series that have risen for a year look 0.9 correlated even if their daily moves have nothing in common.
Series are aligned by timestamp, not by position. A public holiday missing on one side would otherwise shift the whole series by one step and crush the correlation of pairs that genuinely move together.
The practical use is exposure management. Buying EUR/USD and selling USD/CHF at the same time, with a correlation near −0.9 between them, amounts to doubling one dollar position — for a risk that, on paper, looks spread out.