This article proposes a technique for computing sign restrictions in large-scale models. The technique is applied to a Bayesian vector autoregression (BVAR) model with 16 industries (16 growth rates, 16 inflations), and the interest rate. The results demonstrate that the suggested technique can yield different implications for the density of relevant measures compared to the conventional random draw approach. Shocks identification is more accurate for suggested approach in experiments with simulated from DSGE model data. The usage of industry specific data and identification of demand and supply shock have large influence on identification of MP-shocks. It reveals important elements of transmission mechanics of monetary policy including differences in magnitude and shape of responses on MP-shocks, differences in historical decomposition, differences in importance of demand and supply shocks for interest rates dynamic. Variance decomposition shows decrease of relative importance of its own shocks to industries with switching from short-run to long-run decomposition. There are some similarities with input-output tables and some differences those open questions for future researches.