One important point is to gauge the time complexity of the algorithm. This is the most important consideration when deciding how the application will scale.

An example is Ovid's question 1 in (OT) Interview questions -- your response?. Here, a 'solution' to a matching problem is presented in code, but is order n².

Especially where huge databases are involved, getting the algorithm right far outweighs shaving CPU cycles off a calculation. In order to achieve this, it becomes important to think about precisely how an application will scale. If we double the number of customers, how will this affect the run time?