Dear Gemba Coach,
We’re having a heated debate in our company over whether to pursue cost reduction through waste elimination by accelerating kaizen events, or whether to focus on lead-time reduction by implementing a pull system. It appears to me we’re not clear on the link between waste and lead-time. Could you help us clarify this?
This is absolutely the right question to ask and the answer is … overproduction. Please bear with me while I try to clarify why overproduction is considered the worst of wastes, the original lean sin, and how this relates to lead-time. This is a bedrock concept of lean thinking, which needs to be understood in gemba terms.
Imagine a sci-fi gizmo that materializes objects, such as a cup of strong, steaming tea, out of component molecules. There’d be very little waste there: no need to gather tea leaves, package them, store them, transport them, unpack them, process them, throw some away, transport them internally, repackage them in a large packaging machine, throw substandard tea bags away, pack them, transport them to the warehouse, store them, transport them and so on (you get the picture). All the further muda of having people moving around with empty packages, forklifts, and so forth would also disappear. Muda is in fact created by the simple fact that we need to transport materials around in order to produce and then to ship. It’s unavoidable.
Batch du Jour
Now, the equipment we use is neither infinite in its capacity nor is it flexibility. I often argue that kanban is no different to what most restaurants do: the waiter takes an order, which is then placed on a board in the kitchen, and then the cook prepares the dishes in the order they come in. But not so fast. If the cook had to start preparing the soup of the day with every order, no one would ever get served. So a few batches of typical dishes are kept simmering so that when they’re asked for they can be served immediately. And the trouble is anticipating just exactly how much soup you need: too much and there’s leftover (waste), too little and some clients don’t get any.
It makes perfect intuitive sense to overproduce: while I’ve got this set up, let’s make a few more and store them for when someone asks for them. Conversely, let’s group the orders until I have a full load, and then I’ll deal with them. I have to confess that’s precisely what I do with opening my mail or doing the accounts. This is also what we do when we go for the supermarket expedition and why on-line shopping progressively changes everything by enabling more frequent deliveries.
Regardless of how natural it feels, overproduction (producing more or ahead of real demand) creates a number of cost overruns. The first one is fairly obvious: you need far more instant capacity than necessary. This is not always a problem. Laundry, for instance, is done in batches and your own washing machine is used a very small portion of the actual time it sits there. If you were somehow to wash clothes every night for the next day, you’d probably need a much smaller machine (if we knew how to make such a device), not to mention less clothes. Okay, so we like having lots of clothes we rarely wear, and lots of equipment with seldom use. So does the shop floor manager. It’s still waste.
Overproduction is possible precisely because at one given point in time, we have far more manpower and machinery on hand than needed to do just what is required for customers right now.
This disconnect between the capacity really required to follow the sales pace and what we have is the source of most muda. First, you’re over-invested with machines that are unavailable due to breakdowns, slowdowns, changeovers and so on. This may not be felt as a vital problem since we have too much capacity in real terms. All problems remain safely hidden. Second, it’s quite comfortable since overproduction allows for bad parts, missing employees, downtime, etc. However, the drawback is that not only does it hide all problems and doesn’t push you to excel, but it also costs money in terms of more pallets, more forklifts and conveyors and material handling staff, and more warehouses and transports. Overproducing imposes a significant cost overrun on your operations both in terms of capital expenditure and running costs.
Less Inventory, Better Delivery
The irony is that overproducing doesn’t even help to keep customers happy by delivering on time. This is where overproduction links to lead-time. If I’m making several batches of different products on the same equipment or the same line (A, B, C, D, X – where X can be various exotic references), I have to take into account the fact that production time of B is non-production time of A, C, D and X. So the longer I produce B ahead of its sales pace, the greater stocks of A, C and D I will need to keep to satisfy their sales pace. The upshot is that quite a bit of time can elapse before I start producing As again. If I’m changing product once a day, I won’t be producing A again for another week or so. Added up over several steps this creates the lead-time: the time it takes between the moment when you’ve last finished a product and when you’ll get the first one of the same reference from the next run.
This is what you do to customers, and suppliers do to you. When a large order comes in, the supplier pulls everything they’ve got in finished product stock and then puts the rest in of the order in the production queue. The lead-time is thus the time the order sits in the queue plus the time it takes to actually make it.
I routinely find supply chains with 40 days or so in the pipeline – the best ones will be close to a week. This means that a planner has to make a guesstimate of what will be required by customers in two working months in order to create the production program of the upstream link. Not surprisingly, forecasts are often off – and although we have a lot of stock in the system to “protect the customer” we’ve still run out of the specific product they need NOW. When I work with groups of several factories, I usually start by asking them to chart the inventory of each of the plants against its on-time-delivery performance. In almost every case, we get a linear relationship: the most inventory, the poorer the delivery performance; the least inventory, the better the delivery.
Not only is overproduction costly in terms of the waste it generates on the gemba, but it’s also costly in terms of frustrating customers by not fully delivering on time. The link between waste and lead-time is the following: lead-time is generated by production planning decisions, often based on producing ahead of demand (and thus being late on delivery) which involves holding and transporting stuff, and creates waste in terms of capacity and layout as well as hiding quality, personnel and machine upkeep problems:
Lead-time ↔ overproduction ↔ waste
What Toyota taught us is that just “improving” things doesn’t always help much. That’s true when the process is a mess because one can’t understand the cross-impact of what has been “improved”. In many cases, increasing production capacity by increasing productivity (say, from 100 parts an hour to 120 with the same people and equipment) doesn’t really help because all it does is encourage further overproduction. Toyota has taught us that focusing on the lead-time enables you to spot what needs to be improved right NOW.
And so the first question to ask yourself is: when I launch this work order (whatever the process), can I predict exactly when it’s going to be achieved in full (no defects, no missed jobs)? Asking yourself this question will lead you to realize that:
No one believes in matching production pace to sales pace, so the planning is all about somehow juggling the need to maximize the use of equipment and to respond to the customer yelling the loudest right now.
- Batches in the ERP have been set once in a while without good reason and are so ingrained that they can’t be changed, no matter how unreasonably long they are compared to real demand (basically, your production lead-time is your processing time plus your batch size).
- Your process flows are a mess and products crisscross each other in the shop with several intersections and roundabouts instead of freeways, making it very difficult to predict who’s going to arrive when.
- The supply chain is organized by logistics with the explicit purpose of filling the trucks, thus imposing concentration and batching on the entire chain: you need warehouses all over the place because you can’t tell when or what the next sale has just happened.
The next key lesson we learned from Toyota is that by visualizing the flow and the sales pace in physical processes, we create an architecture for kaizen: wherever components are missing or stagnate is crying out for improvement. In this way improving step by step also improves the overall system and its performance. Lower the water in the river and the rocks will appear, which only works if we can see the rocks appear. If the rock is a number in the computer system, we’ll miss a customer delivery and never know which part of the flow actually failed.
To respond to your question directly, there is no debate: kaizen without a pull system will be disappointing. The pull system visualizes overproduction on the gemba, which focuses kaizen efforts on how to limit it, which leads to reducing the cost structure. We just had a bit of fun with one company who’d done so because after a couple of years working with their pull system, they had cancelled the construction of the new warehouse, taken down racks in the existing warehouse, trimmed down forklifts, cut down conveyors and lastly they were selling their huge automated dynamic storage equipment. The fun part is that they were selling them to a competitor who was happy with the bargain price it was getting for prime condition machines. No kidding.
You need to do both. Setting up the pull system is more of a kaikaku project (breakthrough change) than kaizen (small-step improvement) and you need to plan your way to do so.
The first step is to work with production planning to level the demand in order to dampen customer variation in incoming orders. Without this, your pull system won’t be easy to maintain and everyone will end up quickly discouraged. Start by identifying the 10% of your products that account for half the overall volume (in parts, not sales – this is a bet I very rarely lose), then create a plant that assumes we’re going to make the same quantity of these products every day of the week to settle down production. Typically, after having done that, production capacity becomes much clearer and there’s still plenty of time in the day to make all the others.
Then, conduct a project to have each production cell own its finished product stock (and actually see it in front of the cell). This will help you to visualize the link between the leveled demand pace (averaged out through the previous step) and the actual production orders as well as visualize the flow and start to unravel the spaghetti.
Third, use SMED like maniacs to reduce batch sizes by increasing the frequency of tool set-ups without losing too much production time to the production changes.
Fourth, look at the implications on procurement practices and try to level wall-to-wall by increasing delivery frequency and crating level plans for suppliers through the way you order: no scoops!
As you conduct these four activities have one person from logistics work with the kaizen team to work their way through kanban instruction system – it’s not that hard, everyone figures it out and eventually gets to the point of switching off all computer screens on the scheduling system in the shop.
A few months ago, I had the good luck of having dinner with Mr. Masaaki Imaï, who coined the word “kaizen” and introduced it to the West a quarter of a century ago. We discussed the new book he is currently working on and he outlined the key principles of a “kaizen” strategy: flow, synchronicity and leveling. You can’t either/or pull systems and kaizen. Pull creates the architecture for kaizen, and kaizen delivers the cost improvements by getting all minds thinking in the shop floor on how to eliminate the waste revealed by the pull. As my father’s sensei once said: “Never interpret JIT for your own convenience.”