The tendency in any multi-step process that customer demand in an earlier, upstream process is more erratic (i.e., experiences more demand variation) than actual production or demand in the next, downstream process. This is also called the Forrester effect (after Jay Forrester of the Massachusetts Institute of Technology, who was the first to mathematically characterize this phenomenon in the 1950s), the pendulum effect or the whiplash effect . 

The two main causes of amplification of demand at the time orders are received upstream in the value stream are:

1. the number of decision moments at which orders can be adjusted
2. delays while orders wait to be processed and move forward (such as waiting for the weekly run of an MRP system)

The longer the delays, the greater the amplification. This is because more work is then done on the basis of forecasts (which become less accurate the further into the future the forecasts are) and more adjustments are made to orders (by system algorithms that add extra numbers for certainty). 

To minimize amplification of demand, Lean thinkers try to use levelized pull systems with frequent off-take at each stage of the value stream for production and delivery instructions. 

The graph shows a typical situation where the variation in demand on the customer side of the value stream (Alpha) is modest, about +/- 3% per month. But as orders through Beta and Gamma move upstream through the value stream, they become highly erratic; ultimately, the orders Gamma sends to its commodity supplier fluctuate by +/- 35% per month. 

Such a graph is an excellent way to make companies more aware of the degree of amplification present in a production system. If demand amplification could be fully captured, the variation in orders at any point within this value stream would be +/- 3%, a percentage that reflects the true variation in end customer demand.