We present a new method to analyse and reduce chemical networks and apply this technique to the chemistry in molecular clouds. Using the technique, we investigated the possibility of reducing the number of chemical reactions and species in the UMIST 95 database simultaneously. In addition, we did the same reduction but with the ``objective technique in order to compare both methods. We found that it is possible to compute the abundance of carbon monoxide and fractional ionisation accurately with significantly reduced chemical networks in the case of pure gas-phase chemistry. For gas-grain chemistry involving surface reactions reduction is not worthwhile. Compared to the ``objective technique our reduction method is more effective but more time-consuming as well.