Disclosed is a method, apparatus and computer program for sampling a data
feed of messages at regular time intervals. A current sample point is
identified. A time interval is also identified which is defined by the
current sample point and at least one of the previous sample point and
the subsequent sample point. At least one message is selected to
associate with the current sample point. The selection is made from the
identified time interval. The selected message(s) is then associated with
the current sample point.