Photovoltaic (PV) systems are usually orientated to maximize annual energy yield. This may not optimize other system indicators, specifically: direct consumption of self-generated PV power, reduced feed-in power and annual revenue. Also, these indicators are influenced by the energy demand of a building in relation to the PV system size. Therefore, we evaluate how demand patterns influence the optimal PV orientation for self-consumption, feed-in power and revenue. Historical Dutch demand patterns of 48 residential and 42 commercial buildings were used. We combined Dutch and German electricity prices from day-ahead markets with different ratios of electricity sales to purchase prices. Differences between demand patterns caused large variations in optimal PV orientations. On average, PV self-consumption is maximized for residential systems with an azimuth of 212 degrees and a tilt of 26 degrees. Commercial PV systems have an average of 188 degrees azimuth and 17 degrees tilt. Self-consumption can be increased 5.4% for residential systems and 2.7% for commercial systems, by optimizing orientation for self-consumption rather than for energy production. Curtailment losses are significant reduced by decreasing the module tilt angles. Optimizing for revenue can increase annual revenue of PV systems with 5.4% for certain demand patterns and pricing scenarios. The ratio of sales to purchase electricity price has a larger influence on the economically optimal orientation for residential systems than for commercial systems. Differences between Dutch and German market prices have minor effects on PV orientation. Analysed demand patterns significantly affect optimal PV orientation. Therefore, we recommend that optimal PV orientation should not only be based on maximizing energy production, but also on expected demand patterns and market prices. (C) 2017 Elsevier Ltd. All rights reserved.