Convert float to integer simplifies data ArcMap

There is most of the time a loss of information when you convert float to integers. I would say that there are two types of solution to avoid losing too much :

1. rescale the data by multiplying the raster values by a constant (e.g. 1000). In this case, think about the type of integer before you rescale (Byte goes up to 255, uint16 up to 65535 etc). the "optimal" linear rescaling will use the maximum and the minimum values of your dataset: 255 * (rastervalue - min )/ (max - min)

2. reclassify the data into meaningfull group. You will then loose the quantitative information and replace it with categories. e.g. "0 to 3" becomes "small" (stored as 1), ..., and "100 to 1000" becomes "large" (stored as 5).