We have tuned the speech recognition accuracy of many IVR applications and significantly improved the performance by simply tuning speech grammars and confidence levels, with little or no changes to the application call flow. In most cases, we have been able to cut down the speech recognition error rate by around 50%. How do we do it? Come and learn the key tuning techniques that we have developed through the years which have enabled us to achieve such consistently good results.