We propose an importance sampling technique for the conformational analysis of flexible molecules that bridges the gap between currently used uniform random search and random walk methods, and compare performances of these methods in their application to selected open-chain and closed-ring hydrocarbons. It is suggested that, if no information on conformational properties of a molecule is available, the optimum strategy of its conformational analysis should include a uniform random search followed by an importance sampling biased as prescribed by the uniform search and concluded by a random-walk or genetic algorithm routine to rectify the properties of the most important parts of the conformational space.