We measured indoor air speeds generated by ceiling fans in 78 full-scale laboratory tests. The factors were the room size, fan diameter, type, speed, direction (up or down), blade height, and mount distance (i.e. blade to ceiling height). We demonstrated the influence of these factors, showing that the most significant are speed, diameter and direction. With other factors fixed, the average room air speed in the occupied zone increases proportionally with fan air speed and diameter. Blowing fans upwards yields lower but far more uniform air speeds than downwards. We show that for the same fan diameter and airflow, fan type has little effect on the air speed distribution in the region outside the fan blades. We developed several new dimensionless representations and demonstrate that they are appropriate for comparisons over a wide range of fan and room characteristics. Dimensionless linear models predict the lowest, average, and highest air speeds in a room with a median (and 90th percentile) absolute error of 0.03 (0.08), 0.05 (0.13), and 0.12 (0.26) m/s respectively over all 56 downwards tests, representing common applications. These models allow designers to quickly and easily estimate the air speeds they can expect for a given fan and room. We include all measured data and analysis code in this paper.