Wind-driven rain (WDR) is the one of the main moisture sources for building facades. It is an important factor in the dry and wet deposition of pollutants, facade surface soiling and facade erosion. WDR calculations require data records of wind speed, wind direction and horizontal rainfall intensity as input. Most meteorological datasets contain at best arithmetically averaged hourly wind and rain data. Their use is common practice in WDR calculations. As an example, existing WDR standards request at best hourly data. This paper however demonstrates that the use of such data can yield (very) large errors in the calculated WDR amounts and intensities. The reason is that arithmetic averaging on an hourly basis generally causes an important loss of information about the co-occurrence of wind and rain. An improved data averaging technique for wind and rain data is proposed that respects this co-occurrence by applying appropriate weighting factors in the averaging procedure. The performance of this technique is evaluated by WDR calculations on buildings in three cities with different climates. While arithmetically averaged hourly data yield large underestimation errors (Eindhoven, The Netherlands: 11%, Bloomington, USA: 45%, Grahamstown, South Africa: 31%), the improved averaging technique provides very good results (errors: 0%, 4%, 3%, respectively). In conclusion, WDR calculations should not be performed with arithmetically averaged hourly data. Instead, either high-resolution data (e.g. 10-minute data) or hourly data that have been obtained with the proposed weighted averaging technique should be used.