There are a number of ways to measure fitness in biological systems. One method is to partition fitness into components that are standardized across taxa. These components, however, can vary from one taxon to another. Another method is to subdivide fitness across arbitrary taxa. For instance, in insects, survival differences can be broken down into the larval, first, and second instars. However, this approach has its drawbacks.
Normalized fitness is not useful for predicting the outcome of selection
For example, if a genotype dd was not selected, then the overall survival rate of dd individuals would be one thousand percent, and dd individuals would die fewer times and produce more offspring. Similarly, if a genotype dd was selected, the reproductive rate would be one hundred percent, and so on. In this case, normalized fitness would not be useful for predicting the outcome of selection.