An important aspect of any non‐linear inversion method is the generation of a suitable or good initial model as this controls the rate of convergence and accuracy of the result. To overcome the problem, a numerical method is presented for direct interpretation of magnetotelluric sounding data based on the frequency‐normalized impedance (FNI) function. The expressions used to calculate the parameters are developed, first for a two‐layer case under the assumption that deeper layers do not contribute to the early part of the FNI curve, and they are then generalized for an n‐layer situation. The parameters of the first layer are computed by using successive sample values and the final estimate is obtained from the arithmetic mean of selected values by excluding unacceptable results in the logarithmic space. The top layer is then removed using a reduction equation. The repetition of the procedure on successive branches of the FNI function gives successive layer parameters, the resistivity of the substratum being obtained at the final step, when the reduction equation becomes equal to the square root of that resistivity. The proposed method can be used as a complementary method for iterative inversion as it creates an initial guess which is close to the optimal solution. The solution produced by the direct interpretation may also be modified by the interpreter to incorporate prior geological information before being input to iterative interpretation schemes.