Quantifying performance using multiple measurements
In order to overcome the problems of skewing the incentives that measuring performance using single criteria generates, we could think about using many criteria at once. For example, we could see the number of products a factory makes, and the number of accidents in the plant. The problem here is that there'll need to be some normalization.
The biggest challenge is that the way in which we normalize and add together the metrics can change over time and from place to place[@ridgway1956Dysfunctional Consequences of Performance Measurements]. This is exactly what happens when one thinks about the exchange rate of scientific output, since they are many metrics normalized somehow, especially the case that citation-based metrics incentivize competition over collaboration.
Without a single over-all composite measure of performance, the individual is forced to rely upon his judgment as to whether increased effort on one criterion improves over-all performance, or whether there may be a reduction in performance on some other criterion which will outweigh the increase in the first. [@ridgway1956Dysfunctional Consequences of Performance Measurements]
These are the other notes that link to this one.