Or, how to do a pull system in a small batch, high variety environment.
I’ve been working late, but before I go home I nip down to the shop floor for a quick look. As I walk into the machine shop I come across Tim, the QA manager, standing to one side with a vacant look on his face.
“All right mate?”
He looks at me.
I take a step back. Obviously the pressure has been getting to him lately.
Tim laughs and explains that he’s been trying to work out how to do a pull system for our product and had been wondering if delivering components via a train of wheeled pallets would work. I’ve been thinking about a pull system too so we go back to my office and get to work.
Two hours later we’ve got the bones of a workable system thrashed out. It doesn’t involve trains, but it does work well and I’ve used it in a number of companies since. So here it is:
Our product was engineered to order from a mix of bought out and machined in house components. Average customer batch size was about 1.6 so typically items were ordered in ones, twos or threes. Because of the high variety, machining routes varied greatly from component to component and the machine shop operated largely as a jobbing shop. Think pumps or valves and you won’t be far off what we did.
How we organised
We broke each order line down into two kits: a kit of raw materials for machining and a kit of bought-out finished parts. If an order line was for a two off we split this into two sets of kits – ie no batching. When we first set up the system we didn’t start any machining until we had a complete kit of bought-out finished parts and a complete kit of raw materials. Our TACT was 16 finished items per day so we planned to issue 16 kits of raw materials per day to the machine shop.
How we ran it
In the company in question we had a pretty well defined bottle neck process in the machine shop. We put a Kanban in front of this machine (a large CNC machining centre) and used this to pull kits into the machine shop. In other companies with less well defined bottlenecks you might consider feeding kits in at the TACT rate.
Components that were processed on multiple machines were taken by the operator to the next machine on completion of an operation. Occasionally small queues would build up where routes intersected, but these rarely caused problems.
On completion of the last operation on a component it was placed on a pallet next to the bought-out finished kit. Once the last component (almost invariably the one machined on the bottleneck) was placed on the pallet the two kits of components were placed in the assembly buffer and pulled from there into the assembly shop.
Spares and sub-contract
These two are always difficult to manage. Here’s how we did it:
Spares that went through the bottleneck machine were issued in the same way as raw material kits and pulled in via the Kanban. Care must be taken to ensure that these spares are well spread out so as not to dry up work for non-bottleneck machines. Spares that are machined on non-bottleneck machines are pulled from a separate buffer by the operators and used to fill the gaps between raw material kit components.
Some components had an initial machining operation followed by a subcontract operation (eg chrome plating) followed by final machining. These components were issued to the machine shop with the spares for the first machining ops, had the subcontract ops carried out and then issued with the raw material kits for the final operations.
Handling the evil that is set up
In a high variety environment set up is always an issue. We found that the benefits of the way we worked always outweighed the disadvantages of a little extra set up. We also used SMED techniques to drive down set up times.
Supply chain issues
Pushing components into the shop tends to hide supply chain issues. The machine shop is always busy machining components, but assembly is always starved of stuff to build. Working the way we did exposes supply chain issues, resulting in many incomplete raw material kits, but once kits are complete they move rapidly into assembly and so on to the customer.
Product lead time was reduced from 28 weeks to 18 weeks (there was a long procurement time for special castings)
WIP was reduced by 90%
Product could be despatched within one week of issuing a machining kit
Overdues dropped by 80%