Introduction. The article deals with the issues of constructing forecasts in the real estate market using heterogeneous data. Establishing a “fair” price of housing is an urgent task for collateral assigning, for insurance purposes, for determining the investment projects effectiveness, etc. To solve this problem, econometric pricing models are used, which are specified for the entire surveyed population. In case of significant heterogeneity of the surveyed population, the predictions obtained from these  models may contain significant errors. Theoretical analysis. Now, there is a variety of methods and models for the analysis of heterogeneous, spatially distributed data. Population zoning or a variable structure model is used to overcome data heterogeneity. It is connected with a number of problems. An overview of the approaches that implement these methods is given, their advantages and disadvantages are listed. A new method for constructing homogeneity zones, based on the results of the global model estimates building, is proposed to improve the forecast quality. The corresponding algorithm for calculating the local correction factor is described, which makes it possible to correct the global model forecast. Empirical analysis. To demonstrate the effectiveness of the proposed method in action, a forecast for the real estate price, based on the empirical data of the regional real estate market, was calculated, and an analysis of the forecast results was given. Results. The proposed new method for determining homogeneity zones based on the results of forecasts using the local correction calculation makes it possible to avoid a number of problems arising from the use of other approaches and represents an effective forecasting tool.