"And then they are frightened to give it up, because they've spent too much money on it," he said.
"You've really got to get short runs. I like to think most of our projects have a six-month deliverable [and] if it doesn't deliver, be prepared to chop it off at the socks rather than keep putting money into it."
Knowles said trying to fix problems can be like the gambler's problem - "I'll put another bet down, because I might cut my losses [but] eventually you go bankrupt."
The concepts of source code access and not being reliant on one vendor for support are valid, according to Knowles, but a decision to deploy open source comes back to risk management.
"Open source in a lot of cases may well be the low-risk strategy, but you need to be careful of where you are going with it," he said. "You've got to say 'would I go and develop a financial management system or will I work with an SAP [AG] and so on.'"