Data collection and data interpretation have become main issues in modern information society. Concerning this matter preparation and visualisation of data are essential aspects for any analysis. With an appropriate illustration one will get easy access to the underlying information of data. Therefore, the chosen illustration method can have a major influence on the interpretation of the data. For data containing geographic references, map presentations are frequently used as they are the best method to illustrate spatial relations with the help of coloured symbols, borders and areas. The data discretization and the colour codes for the categorisations may result in huge differences of its visual impact. When dealing with aggregated data the approach of pre-processing becomes a key issue for obtaining informative maps. To get comprehensive information for all geo-coordinates of the region of interest a new non-parametric approach for density estimation named “kernelheaping” is applied and evaluated. Based on the election results for the German Bundestag in 2017 the new technique is compared against standard choropleth maps and some modifications, like normalisation. The iterative kernelheaping procedure is also compared to non-iterative variants of kernel density estimators. In addition, the new approach is evaluated statistically with an underlying known real-world density on different aggregation levels. The Master thesis was written in cooperation of the Statistical Office of Berlin-Brandenburg and the statistical department of the economic faculty of Freie Universität Berlin. Beforehand the author completed an internship at the office. In this internship the kernelheaping method was set up and tested on preliminary election data and subsequently put into practice on the election night by using real time data. The results of the regional analysis were published in the official election report and the journal of the Statistical Office of Berlin-Brandenburg.