I use 100% incremental multiples of your target size and make the first one the "small" one. So for example, if your target size is 900 pixels on the long end, and your camera is 5000 pixels across the long end, I suggest you make your first downsize to 3600 pixs. Then go to 1800, then finish at 900. I use sharper for the first and last, regular bicubic for all middle steps. This method is it is easily actionable to get to target resolutions. If you do add a "fit to" step at the end to keep vertical and horizontal images the same max size. I happen to have a set of 7 web-converter actions pre-packaged -- a "full size" and then a 1200, 900 and 600, one set for high res cameras (20MP and over) and another for medium res cameras (10 - 20MP).
Cheers,