So supposedly you have
- tested the properties of the two time series
- passed cointegration test
- and you are sure that the correlation in this data period was not driven by a very obvious macroeconomics factor, then it comes to defining your investment strategy.
A common one is pairs trading, where you remain cashless in longing one asset and shorting the other. There are lots of ways to do pairs trading, but as a starter let's say you defined the spread $\epsilon_t $ between asset A and B (counting the ratio mentioned earlier to maintain cashless), since we think there's long run equilibrium between the two asset, $\epsilon$ series should follow normal distribution like a noise.
Your buy/sell signal then can be: once the spread exceeds say 2 standard deviation, we think the current spread is overvalued and thus we short the asset as the subtractor in our equation defining spread (e.g $\epsilon$ = A- some ratio * B, in this case asset A) and long the other one, vice versa on -2 std.