How Much Contingency Is Enough?

My best guest generally 5%

• There are a number of ways of calculating contingency, the most common are; Hidden Contingency, Project Contingency, and Phase Contingency, those mentioned in the referenced link article. These are three ways we have seen in the past. For some they make it for others there are better ways to deal with contingency.
• We consider inflating activities duration a bad practice, you will never know how much buffer is considered and what is the un-inflated activity duration. Inflating activities duration as well as buffer activities will also delay projected early start of activities in the future that might start earlier if less than 100% of buffer duration is consumed, this is misleading. As the author of the article said you shall plan in a way not all contingency is used so it is to be expected not always all buffer will be fully consumed.
• We have a preference of statistical methods such as Monte Carlo because the other methods fail to recognize that usually there are many sub-critical activities belonging to the different network paths and even the minor delays in the execution of sub-critical activities can lead to changes in the RCP [Resource Critical Path]. This comes into conflict with the Critical Chain theory’s assumption that the Critical Chain never changes during the project execution.
• For day to day operations we use a tight schedule with no buffer. We plan for a tight schedule with enough resources that will yield good probabilities of success for meeting contract duration and contract milestones. 
• For contract management we use a buffered schedule using a combination of any of the three methods we do not find suitable for operations management but viable for contract management. When statistical methods such as 3 scenarios or Monte Carlo are accepted we favor either of these options as both are embedded within the software we use, Spider Project. In any case we plan our buffers using statistical methods in most of our jobs, arguably somewhat arbitrary but better than fully arbitrary contingencies.
• http://www.spiderproject.ru/library/cchr.pdf

Hi Emily

How much contingency is enough?

As much as you can get away with.

Best regards

Mike Testro