Branches of an insurance company were measured on workstate returns (backlogs). Reading branch consistently showed a nil backlog. Why? They put all the outstanding work in the post (to themselves) on the day the measure was taken.

Zero backlog; best operator. Reading had the worst service.

Often when you study backlogs you discover that the same customer can be responsible for many ‘items’. For example, if one customer transaction accounts for seven items in the backlog, then six of those items may be failure demand. Think about it: managers bring resources to bear on backlogs, but maybe the backlog is inflated – not a real measure of ‘workstates’.

Managers use measures of workstates to move resources (people) around to deal with the work. They think they need to ‘hose down’ backlogs and/or large queues of work. Moving resources around in this way is a form of tampering; it makes performance worse.