(excerpted from The Pipeline; Part I was published on Escape from Averageness in October 2009 and Part II was published in March 2012)
“Our production system is a pull system, but that wasn’t always the case”, the CEO said. “The current system – which is new – replaced a push system. Like the current system, the old system had a start matrix, too, but it controlled both the rate and order of starts in a community. The starts slotted into the start matrix were regarded as untouchable – once they were scheduled, starts were not supposed to be missed under any circumstance.
“It was noted that the entire building process is protected by a buffer of available starts, which protects the start matrix from fluctuation in sales with an “inventory” of starts. That’s what we call the start buffer, and it actually sits in front of the start-to-completion process.
“Every other homebuilder proudly calls this their “sales backlog”, as if having a six-month backlog of contracts that they can’t start is something they actually want. Under the old system, we looked at the backlog that way, too – the longer the backlog, the better. We would sell a house in April, with no intention of starting it until October. Now, we take a different approach.
“If our controls will allow the start, we certainly don’t want to miss it, but – over time – we concluded that a start buffer with a 60-day backlog provides sufficient protection for the start matrix, and corresponds to the current length of the contract-to-start process. Any lengthier backlog – a start buffer that’s any larger – is muda; too much to manage, too much that can change. Our homebuyers hate the wait, and it really is counter-productive.
“Under our old system, the matrix did a good job of producing starts at an even rate, and gave needed order to what was previously a totally chaotic process, but pushing starts into the system without regard to the throughput – the rate of closings – resulted in higher levels of work-in-process. Absent any increase in either productivity or production capacity, RB Builders’ cycle times would lengthen, to the point of sometimes reducing the rate of closings.”
The intrepid, results-based consultant picked up where the CEO left off.
“That’s a good explanation”, she said. “For most of us, once we’ve had a chance to think about it, “pull” makes immediate sense, because it is linked to the demand that justifies the production. But, the issue of capacity – whether it should be balanced or unbalanced – is less intuitive.
“It is a paradox. It is a paradox that a production system with balanced capacity cannot achieve an even rate of production. The term “balanced capacity” means a production system that has its capacity distributed evenly throughout the system”, she explained. “It means a production system designed with the same capacity at every resource.
“A production system with the same production capacity at every resource – in other words, a system that purposely levels its capacity across all of its resources – cannot have even-flow production. Intuitively, we believe the opposite, that a “balanced” system – one in which resources have the same capacity – produces balanced results. However, in a production system with balanced capacity, variation and uncertainty anywhere in the system will affect production everywhere in the system, making it impossible to control or predict.
“Moreover, production systems with balanced capacity tend to be very rigid and difficult to manage. They are not the adaptable, agile, easily-managed system we want RB Builders’ production system to be.
“Even-flow production – which we want – is an outcome, not a mechanism. In order to have an even rate of production, we have to purposely unbalance the system that produces it, and create production “pull” instead of production “push”. We have to live with some amount of excess or reserve capacity on the non-constraint, on the non-pacemaker resources.”
“Doesn’t this fly in the face of Lean and TPS?”, asked the VP of Construction. “If we want to embrace Lean Homebuilding, don’t we have to do heijunka – don’t we have to do production leveling?”
“In a controlled manufacturing environment with continuous flow, production leveling works”, said the intrepid, results-based consultant. “The plant can move equipment, cross-train workers to do other workers’ jobs, shift production cells, change product mixes, adjust production runs, etc. Problems can be solved much more rapidly in that environment.
“However, in an environment like homebuilding, heijunka is exponentially more difficult to achieve. Homebuilding is not a controlled manufacturing environment. It is the equivalent of building cars in people’s driveways. Instead of teammates, we have trade partners that are independent sub-contractors and suppliers, who also work for our competition. We deliver materials to hundreds of jobsites. The more inherent variation and uncertainty is to a process, the more difficult it becomes to level production.
“Moreover, homebuilding is not the continuous, single-piece flow process that Lean prefers. It is a build-to-order process”, she said, motioning for the SR Chorus to keep their seats. “Lean gives you its first clue about the feasibility of production leveling in a build-to-order process, by recommending that the most constrained resource – the bottleneck – become the pacemaker. If the capacity and production rate of every resource is leveled to the capacity and production rate of the constraint, then, theoretically, there is no constraint. And – theoretically – you would have unlimited capacity.
“Yet, we know systems do not have unlimited capacity. There is always a constraint. There is always a weakest link in the chain. We are far better served purposefully placing the constraint and subordinating everything else to it, than we are fighting for a system with balanced capacity.”