Should we plan according to what we can do, or the other way around?

Dear Gemba Coach,

Should we plan according to what we can do, or the other way around?

The big idea behind takt time is that you adjust production plans to demand, rather than plan delivery according to production requirements. A core idea in lean is that sales pace = production pace. At the gemba this is often impractical, for a number of reasons. First, sales volume tends to vary from one day to the next whereas efficient production requires some degree of stability. Second, even if the overall sales volume is relatively stable, sales mix will vary from one day to the next, and production is rarely flexible enough to deal with such variations.

In assembly, for instance, one line or cell should be able to make any product at any time, and ideally follow exactly the sales pattern: one sold, one in the queue, one made. But when you go down to the shop floor, you find out that’s rarely the case. Most production cells are set up so that every product change requires emptying the component racks of the parts for this product, and refilling the racks with parts for the next product. There is no tool change as such, but production change is still disruptive because of component changes.

Years ago, I once asked a Toyota engineer why he was insisting so much on really small containers at the workstation – I couldn’t see what the possible gain in inventory was at the level of halving a container that was already quite small. “It’s not about inventory,” he answered, “but takt-time. Ideally I want to produce products exactly according to takt time, which means that I wont produce two same products in a row. The operator needs to be able to switch from one product to the next seamlessly, so they need to have all components of all variants at all times in front of them. Since space is limited at the workstation, this implies very small containers to fit them all within easy reach of the operator.”

Pressed to Be Lean

Similarly, one step further on the value-stream map (VSM), at the press, changing tools at every product is simply absurd. Assuming we’re willing to invest 10% of Overall Equipment Effectiveness (OEE), which is already a lot,  in tool changes, a 15 minute tool change implies batches of 150 minutes of production (roughly, ten times the changeover time). If the press makes a part a minute, this means a batch of 150 parts of the same product. In many plants, press changes are closer to 45 minutes than 10, so it’s not unusual to find batches of 450 minutes (a full shift) or of an entire day.

Say the process is relatively simple, as with one press operation and then a simple assembly of adding a few components to the pressed part. Let’s assume that one press handles two high-runners, A and B, and three or four low runners, C, D, and E. After years of wrangling with the press shop about reducing batches they’ve accepted no longer to run the high runners once a week but twice. Allowing some time for changeovers, they run As on Mondays and Wednesdays, making 2,400 parts in six shifts, Bs on Tuesdays and Thursdays, making 2,400 parts as well in six shifts and make the X part (C, D or E) on Friday in three shifts – overflowing on Saturdays if need be.

In this situation the argument will be to plan the entire value stream (press + assembly) as the following:

Lean? Not so. Regardless what the press shop’s flexibility is, the demand for A is 2,400 parts, for B 2,4000 parts and, for C 400, D 400 and E 400 parts as well. So if we level the demand across the week, we end up with the following production plan:

Total daily demand is 1,200 parts whereas the press has a capacity of 1,350: 150 minutes are available for changeovers. If tool changeovers are 40 minutes, they should just about make it. But the moment one changeover takes, say, an hour, they start losing parts and won’t be able to fulfill the demand – plus all the extra hassle of doing four to five changeovers a day instead of one.

Press shops being what they are, whatever the production plan, they’ll stick to what they’ve agreed: run 1,200 parts of A on Monday, push them into the WIP warehouse and let assembly do what it will.

SMED Metric: Rabbit’s Foot or Chicken Leg

This is where lean kicks in. In the future state of the VSM every process owns its production. The press owns its inventory, so the parts hotel in the warehouse is shut and the press gets to keep all parts it produces. In this case, what the production plan tells the press is that it will come and pick up 480 parts of A, 480 parts of B, 80 parts of C, 80 parts of D, 80 parts of E at some point during the day according to how kanban cards flow in the pull system – regardless of how the press shop chooses to schedule its production.

This lean production plan doesn’t tell the press shop what to produce. It lets the press shop know what its responsibility is. At the press’ supermarket, the press shop has to make this number of products available every day of the week. Now, clearly, if it decides to produce 1,200 parts in one run, it means that at the end of Monday’s third shift, it will own 720 parts of A in inventory – and its parts supermarket needs to be large enough to accommodate this.

As you can expect, this will be a big fight as well as create the incentive for the press shop to get its changeover time down to the level where it can produce every part every day.

In fact, even producing once a day makes you carry huge inventory. Let’s consider that by improving the assembly line’s workstation, we can now produce parts in batches of one half hour. The pull system’s tugger can run through the press shop to pick-up parts accordingly every half hour – which means 45 pickups across the day’s three shifts. Since A and B are the two high runners, the train is likely to pick up some A every hour or so, and some B as well.

To follow the train’s pull ideally, the press shop should produce in 20 runs of 60 parts. This would mean 20 changeovers to be done in 150 minutes: 7.5 minutes per change over. SMED (single minute exchange of die) legend Shigeo Shingo is said to have walked through press shops with a rabbit’s foot and a chicken leg in his pockets. If the change-over was below the 10 minutes line (single digit), he’d pull out the rabbit’s foot. If not, you got the chicken’s leg. Changeovers of 7.5 minutes are not impossible – I’ve seen a 4,500 ton press, as large as a house, changed in five to seven minutes in Toyota’s stamping shop. But it’s quite a challenge. It’s lean.

All this to say that the leveled weekly plan is established according to averaged customer demand day per day. We average customer real demand over three weeks, and then we divide this by the number of worked days. This gives us how many parts will be picked up from the supermarkets in the VSM any given day. The steps in the process may not be flexible (or willing) enough to be able to build at that rhythm. They might have to build once every couple of days, or even (shock, horror) once a week. No matter. The shop’s abilities should not restrict the production plan. What is planned is their responsibility for the number parts to be made available on each given day. How they do it is a different debate.

Legend has it that Ohno’s last words were “no kanban.” Until all steps in the process can produce according to a perfectly leveled, fractioned and mixed schedule, we will have to batch, and so, to pull. Our capacity to do so should not affect the plan. On the contrary, the leveled production plan is there to tell us what we need to have available in our little cell’s shop, every day.