This paper describes a method that can handle the uncertainty in the VAT data at the time of the GDP flash estimation. As the VAT database in Hungary will only be finalized six months after the reference year, the expected revisions cause a problem in the estimation. Using historical VAT data of all available VAT submission, a vintage matrix is compiled in order to forecast the revision of VAT data with sate space model. Based on the result of our model, the quality of growth rate is significantly better.