Recently, there has been an increased interest in the scaling behavior of stock returns. We propose a new wavelet-based method for identifying and characterizing the multiscale dynamics of nonlinear financial time series. We have developed a wavelet framework for summarizing multivariate financial data with "chief features" like trend, seasonality, turning points, changes in variance and systematic risk. Our framework can generate quantitative parameters (like slope of the trend, period of seasonality, and regions of low and high volatility) to characterize these "chief features". Thus the summary output from our framework describes the market dynamics at various timescales and can be used to generate "buy / sell" signals and predict future returns. We report results of experiments performed on the Dow Jones Industrial Average (DJIA) and the 30 companies within it. Notably, the wavelet analysis appears to be a potentially powerful technique for assessing the multiscale dynamics of financial processes. We conclude that the development of an efficient wavelet-based prediction model that compares well or outperforms other established models like the GARCH model is possible.