Manufacturing a log cabin

Recently, as part of a project, I visited a metal parts supplier for larger integrators. On the factory floor, the head of the planning department explained to us how orders were planned and tracked in production. He also showed us the progress of certain orders and how many of them were being processed. The numbers surprised us. There were much more orders in progress than we expected, and the lead times (up to 2 months) struck us as well – given the size of the orders and the kind of product.
What surprised us most, however, was a traceability issue. The planning department knows exactly which orders are in progress and when they’re due. They also know how much of an order has passed each production step. What they don’t know is when the order is going to be finished. Note the difference: they know the due date but not the completion date. The two are closely related and quite crucial to meeting customer agreements.
 
Baby steps
This factory isn’t under fully automated control yet. Like many small production facilities, they’re interested in smart-industry ideas. At the same time, they don’t have the time, nor the budget to automating in one big step like huge manufacturers such as, say, L’Oreal or Volkswagen. They’re making their first baby steps and we’re helping them.
The problem we unearthed here can very easily be explained: the planning department is new and still working out how to plan and track the complete process. Right now, the individual departments (or manufacturing steps if you will) each plan their work on a day-to-day basis, using due dates and machine capacity as starting points. Every day, they plan parts that have the closest due date, and then fill up with whatever later order they have lying around. With due dates as far as six months ahead, large orders are initially mainly used to fill up spare capacity, until they have the nearest due date and get priority.
As a result, some parts are produced in week 2, some in week 8, and so on. Sometimes they’re finished before their due date but also quite often the due date isn’t met – because of this way of working.
 
Uncontrolled flow
To get the issue clear for everyone involved, we set up a metaphor. Assume we have a river that flows into the sea. There are also smaller bodies of water feeding into the river itself. Now we want to build a log cabin at the sea mouth, using logs made from trees carried over the river. These trees fall into the river when they break off in a storm or at the hands of a bunch of lumberjacks we hired. These guys are up in the mountains and they like to party, so sometimes they chop a lot, celebrate and then, being hungover, don’t do anything for a few days. As a result, trees come in at our building site at random intervals and in random numbers, making it nearly impossible to build our house. The only thing we know is that the lumberjacks get paid for completing their job before winter and they have to come back down and wait for spring to go back up again.
Such a way of working means that days, sometimes weeks, go by without us having anything to work with, followed by an overflow of material. Still, we have no guarantee that we can finish building before winter sets in on the coast. This describes, in a way slightly more black-and-white than reality, how our factory is currently operating. The planning manager understood and asked what he could do about it.
We then explained that what we described is essentially an uncontrolled flow process. To get more control, we suggested switching to batch production: split orders into batches of an equal number of products and completely plan the process for these batches, with start and end dates per batch for each department.
 
Batch planning
We explained our solution again using our log cabin metaphor. Imagine more and more people getting interested in building cabins on the coast – may be because the original builder has raised a family or maybe other people moved there. Building from randomly arriving trees isn’t a feasible way anymore to make sure everybody has a roof over their heads before winter, so some adjustments are made in the lumbering and building process. The lumberjacks get a contract specifying how many trees per week they should cut and send down during summer and autumn. Partying is moved to Friday and Saturday evening and recovery to Sunday. Also, as trees sometimes got stuck for a while behind rocks in the water, the lumberjacks are told to bind them together into rafts. Each raft is manned by a ‘gondolier’ who keeps it from getting stuck on the way down.
We now know on a week-to-week basis what comes in and can plan our building process accordingly. Note that there may still be rafts arriving a day early or late, but on average, we’ll get it right. This is what batch processing and planning are about.
At the moment, we’re working on a plan for this factory to further improve and automate its processes. Introducing batches will be one of the first, but likely not the very first step.
 
Nothing new
It seems like a trivial thing to do, but we already noticed the same issue at different factories. This shows that getting data out – production amounts and lead times in this case – is a useful contribution to process improvement. It’s nothing new, though. Eli Goldratt, the author of books like “The goal” and inventor of the Theory of Constraints, already taught this in the 1980s, but apparently, there are still many places where this is not the way of working.
Surprising, but also challenging, and we’re up to the challenge. Time to make Shinchoku – progress.