A time buffers placed in project chains (paths) that tie in to the critical chain. Their purpose is to help insure that the successor task on the critical chain has all of its inputs with about a 50% probability.
Size the feeding buffers based on the tasks in the feeding chain. The buffer size is the sum of a buffer for bias plus a buffer for variation. (Advanced Projects Institute)
Other Definitions
Feeding Buffer – Feeding Buffers protect the ability of the critical chain to maintain its relay race performance by buffering the variation of non-critical tasks and chains where they feed into or merge with critical chain tasks. With a properly sized feeding buffer inserted, the critical chain task that relies inputs from that non-critical chain has an improved chance of being able to start as soon as it predecessor critical task is complete.(Frank Patrick’s Focused Performance Critical Chain FAQ)
Feeding Buffer – Is the extra time provided for activities on the non-critical path, which merge with the critical chain. The feeding buffer provides a guard against delays in non-critical paths affecting the critical chain. The feeding buffer, as a thumb rule is normally assumed half of the time removed from activities on the feeding path. (MangementMentor.com)