Absolute and Relative Performance

It is imperative that any performance measurement has to be done relatively among comparable samples.

Let's take a very macro view of this statement, how do we compare economic performance of countries? Malaysia, for example annual GDP growth adjusted by inflation for year 2008 was 4.73% on average and the same measure for year 2009 was
-3.77%. Looking at these numbers by their absolute sense only be able to tell you one-sided of the story, the magnitude of change between contiguous time periods. How do you tell whether 4.73% for year 2008 was actually top performing achievement and its fall to -3.77% was in fact within acceptable normal range? The answer to all these questions are not easily obtainable. To compare and interpret the numbers, we must have data of countries of similar characterics which include developing status, government types, national focuses, comparative advantages, currency power and so on.
Unfortunately (fortunately), this planet only logically partitioned into less than 200 countries (officially according to UN) and this means that the samples that fulfill the suitability criteria can be significantly little.

Performance must be measured in relative term! It makes more sense to state that Malaysia is performing at 95 percentile of all qualified samples or Malaysia is performing 5 percentile better than Thailand.

Applying the same analytical process to evaluate a group of employees or set of business units where each consists of a group of employees, it is not likely that neither employees are all performing nor they are all not performing. Performance might be normally distributed and it implies that most (67%) employees' performance fall between 1 standard deviation from the distribution mean. With more sample points, I do believe that this will be the true population distribution, however with less number of employees in the analysis, we perhaps can observe skewed distributions, positively or negatively. In layman terms, we could observe majority of the employees are performing exceptionally good or worse. Taking a negatively skewed distribution of employees' performance, what will be a suitable compensation system? Should we lucratively compensate exceptionally performing people that only registered themselves near to the distribution mean or severely penalize people far at the left skirt of the graph who obviously fall on the definition of outliers.

No easy answer because it is impossible to have a undisputable and universal performance benchmark on human activity. It is difficult because even such benchmark and measurement methodology exists not everybody will bow down to its results nor agree to the rationale of every measurement.

I guess people management is undeniably distantly far to be an engineering disclipline.