Saturday, June 2, 2012

To pull or not to pull… the German Autobahn would work much better if it would be controlled by conWIP.


A production line is like the German Autobahn. If you keep pushing, it will break down. Like a production line, the Autobahn has measurable parameters. Release rate; at what frequency are cars coming on to the road from the ‘Einfahrt’. Urilization; what is the percentage, or portion, of the road that is occupied by cars. Variability; to what degree are drivers out of sync with the speed at which they are driving, slowing down, stopping and starting up again. Lead time; how long does it take cars to travel a certain segment of the road. And throughput; at what rate are cars leaving the Autobahn through the ‘Ausfahrt’.
So what happens when the Autobahn underlies a push principle?... let’s take a step back here and first explore the question: “What is it, that implies we are working within a push principle?”
In a push environment, entities enter the system according to a planned or random moment in time. A planned order released to the production line according to the planned start date, a baby entering this world (there’s your push), a car driving onto the Autobahn. All of these entities enter the system without any regard of what’s going on inside the system. A newborn does not care what the world is like; whether there are too many of us; it just wants in. A car going on a trip is a bit different (some people might listen to the traffic report), but not much. And your jobs scheduled to go on the line? Are they not being released into the process when the packaging line at the very end is down? Of course they are. Otherwise the plant controller would scream murder, when with the packaging line, the other work stations go down too.

But herein lies the difference between push and pull. In a pull environment, the release rate is controlled by the system itself and what goes on inside. Take air traffic control; If the runway for landing in Newark is down, every plane within the system that wants to fly to Newark is placed in a holding pattern and therefore WIP, or the inventory of airplanes in the air, is going up and it starts getting crowded around the New York airspace. Some smart person has, at some point in the past, decided to put a limit on the amount of airplanes that are allowed to fly around, within the New York airspace. Once that limit is reached, no other plane is allowed to take off from any airport to go towards New York.

I call that the WIP cap and that is what makes the air traffic control system a pull system. In a pull system, once the WIP cap is reached, another entity is only allowed to enter, if another entity leaves the system, so that there is constant WIP.

Let’s summarize: A push system is driven by start dates; a pull system is controlled internally by way of a WIP cap. (That also makes clear, that babies will always be pushed into the world)

Back to the Autobahn. Back in the 60s when there was no traffic radio or iphone traffic information, the Autobahn was a pure push system. Later, all these traffic reports were introduced and we control a little bit better, but not much, when cars will enter traffic. I we were able to enter the era of pure pull in automobile traffic, everybody wanting to travel, would receive a spot at which time enter the road. In the perfect world you would stay at home until your ticket is called and then you would enter the Autobahn at the same utilization (amount of cars per kilometer) as always. (btw… if we are able to extend the air traffic systems so that you get a call when to go to the airport, we could avoid the piles of passengers at airports when there are flight delays).

All of the above is very desirable and people are working hard to make this a reality sometime. To switch a production line from push to pul is far more easy, but it still doesn’t happen much. Why is that?
In my mind it is because we do not point out the inefficiencies resulting from push with a scientific method like Little’s Law, the VUT equation (by Factory Physics) or a value stream map. We just look at the perceived cost of having a work station down. There are other reasons to not release a new job onto the system (e.g.waste of overproduction) and when WIP piles up downstream, it is certainly not smart to keep introducing raw materials to the process upstream. But there is always someone shouting that the lines must be utilized as much as possible.

We have to prove these people wrong!