Home > Community> Gemba Coach> How do I re-size supermarkets when demand changes or to keep pace with seasonality?

How do I re-size supermarkets when demand changes or to keep pace with seasonality?

Michael Ballé
Permalink   |   5 Comments   |   Post a Comment   |  
  |   RSS

Dear Gemba Coach,

How do I calculate re-sizing supermarkets when demand changes or to keep pace with seasonality?

There are two very different ways to look at this. The most common perspective I see on the shop floor is SAP thinking with lean tools. In this sense, the question is: we have calibrated our supermarkets and kanban cards for a certain level of demand, at which point has the demand changed so much that we need to recalibrate. The assumption here is that the pull system should work the most effectively way possible and tinkered with when conditions change. In this perspective, kaizen is necessary to keep the kanban working.

And then there is the path less travelled: using kanban as a tool for kaizen. It’s a question of perspective. Let me try and clarify, I’ve recently come across a 1970s story about Taiichi Ohno visiting a Toyota supplier to whom the company was buying seventy thousand parts per month. The plant manager tried to convince Ohno the company had enough capacity and manpower to cope with 100,000 parts if need be. “Then,” Ohno asked, “do you close your operations ten days a month since we are only buying seventy thousand parts?”

“We wouldn’t do anything that silly,” the man answered. “We are building a warehouse for the excess production.” Ohno then explained that if he built the warehouse he probably would lose his contract to Toyota since the additional overhead would make his parts too expensive. The idea, Ohno said, was to have only the equipment and workers needed to produce what was actually sold. Future production increases should be obtained by kaizen. (Against all odds, Togo & Wartman).

Plan for Every Part

Although predictable lean gospel, this story struck me (the book is about Toyoda history, and this is the only mention of TPS) because it goes to the heart of the matter and it is still as relevant today as it was when it occurred in 1974. The size of a supermarket to make the flow work was never Ohno’s concern. His unique perspective was a keen sense that any unnecessary overhead added itself to the cost of making each product. In other words, it’s easy to resize supermarkets when demand increases – just increase them. Or when demand decreases, simply let them as they are.

Back at the gemba, I know a service operation that needs a stock of service parts to supply maintenance technicians over a wide area. In the first year of doing lean, they divided the parts inventory by three mostly by isolating obsoletes. They realized in doing so that a small percentage of parts were used frequently out of the total of thousand possible parts in the catalogue. Without changing anything drastic, they installed a supermarket for high runners, which improved service OTD and reduced inventory. But then what?

The next step was to go all the way to the plan per part, a simple spreadsheet in which each part in the supermarket is defined by the vendor, minimal purchasing quantity, delivery lead-time from vendor and price. The vision was a sort of Amazon same-day delivery service: how could they purchase parts as they were needed by technicians so that they could be received singly in one day at a competitive price?

What came out of this hard work was a realization that the company wasn’t too good at:

  • Estimating technician demand – the technicians often changed their mind while in the field about the number of parts they needed for one component, which made it difficult to procure and get it to them;
  • Long lead-time from suppliers – some orders to vendors could take weeks before getting to the warehouse;
  • Large packaging – many traditional industrial vendors would only sell the parts in large packages, which made the cost per part ridiculously low, but also created months of inventory of one part at any one purchase. This means that the cost per part is lower than purchasing a single part elsewhere, but the cost of the entire packet is much higher than a single purchased part (even though single purchase will be expensive), and the n-1 parts will sit forever on the shelves.
  • Bad parts and other mistakes – typically, when one opens a large packet a high number of parts in there are of dodgy quality. When one supplies parts one by one, the reactivity on quality is much better.

TLC for Inventory

The insight from this exercise was that the inventory was a living thing that needed tender loving care. So rather than adjust it when demand varied, the guys decided that they would re-simulate their inventory once a month using the plan by part to better understand how it behaved, and to learn how to better deal with specific items. The radical change of thinking here is that the monthly simulation of how the inventory behaves according to the expected demand conditions drives specific kaizen topics:

  • How do we better understand field demand (there is seasonality), and do we have a plan per every equipment in the field?
  • How can we get closer to a instant part by part supply?
  • How can we further reduce the inventory and at what immediate price?

There is no set method to recalculate the supermarket of high runners. There is a monthly scratch-your-head exercise of looking at it line by line and wondering what we’re going to try with each part. There is no set formula, no function of lead-time and packaging size precisely because we want to challenge assumptions about procuring each component, not apply across the board formulae.

The same thinking, I believe, applies to WIP supermarkets within the flow. The reason the supermarket exists in the first place (with its related overhead) is the gap to true one-piece flow:

  1. Demand surprises, where the real demand is suddenly higher than expected and we’re short of parts, or when demand plummets and we don’t know what to do with the parts on hand
  2. Insufficient machine uptime (whether because of equipment or labor availability problems), making it necessary to produce ahead of demand
  3. Slow change overs, making it difficult to shift resources instantaneously from one component to another
  4. Large batches that multiply the lead time to get a component since time of production of A is non-production time for B, C, D and etc.
  5. Etc.,

Internal supermarkets are regulated by kanbans, so the previous thinking applies: every month we look at part-by-part behavior and simulate how the kanban generates this behavior with the twin goals of no missed-deliveries while reducing the overall inventory.

Here again, there are several formulae to calculate the link between kanban cards and parts, and some are useful, but that’s not the point. The point is taking kanban cards out, reducing the supermarkets and identifying the resulting kaizen projects, such as:

  1. Better understand demand (and seasonality is part of this, we treat seasonality as a special customer which we can level through the year)
  2. Reliability of equipment in terms of quality and uptime
  3. Flexibility of equipment
  4. The internal logistics or running trains and issuing kanban cards

In other words: the usual suspects. To your question, I would not advise you to try and find a standard method to resize supermarkets according to demand changes or seasonality.

Lean Is About Leaning

I would set up a demanding monthly review board to constantly challenge how each supermarket works and push to reduce them (sure, in exceptional circumstances, we can extend supermarkets to fend off a crisis, but that is profoundly bad!). One of the main temptations of any pull system is simply to replace the existing MRP with a supermarket/kanban mechanical system without changing the underlying thinking. Lean is not a state, it’s a process: you’re never lean, you lean what there is.

Too many manufacturers put in supermarkets, kanban and trains, and think, “Hey, we’re lean now, what’s next?” Lean is about leaning, which is about having the discipline to consider the question you ask seriously on a regular basis, and use the size of supermarkets as a driver to kaizen deeper technical questions such as our understanding of customer behavior, our understanding of our technical skills and our ability to improve flexibility. There is only one way to resize a supermarket: cut it by half!

5 Comments | Post a Comment
Phil Coy November 14, 2013
"There is only one way to resize a supermarket: cut it by half!"  Really??

I was tracking right along with all of your thinking till I starting hitting the statements that culminated in your last line.  Still I agree with your thought that kanban is a tool for kaizen and a relentless push to reduce inventory is useful to surface underlying issues. 

What I don't agree with is your assertions that there are no formulae, that you shouldn't use them, and that you need to deeply consider every item on an item by item basis and make an individual judgement call.  I hope this is hyberbole.

I work with high mix/low volume manufacturers.  The company I am currently engaged with has 190,000 items.   They have seasonality and are configured make to order.  That's a lot of parts for individual subjective consideration.  Just "increase them when demand goes up" or "leave them alone when demand drops" leaves me cold.  How much do I increase them?  Is demand going up because it's early in the product life or because of seasonality?  Is demand dropping because of new products or just the slow season?  Why should I replenish to a certain level when demand drops?  Isn't that leaving waste unaddressed?

There should be reasons for the level of inventory chosen that should be documented.  And there should be a quantitative understanding of the risk the company is willing to bear for stockouts when (not if) demand exceeds any planned level of buffering.  This may sound like "SAP thinking" but in truth supermarkets are inventory buffers to cover the variation in demand.  Inventory levels to avoid stocking out can be calculated based on the capable EPEI of production loops and processes.  It's "lean math", not outmoded MRP or unintelligible APS/FCS.

My observation over the past few years is that lean is moving away from its roots as quantity control as referenced in some early Ohno statements.  While the tools alone are not sufficient for sustainable transformation, the tools are essential elements that can't be ignored.  Otherwise lean becomes another management system alone and the power of the lean tools is lost.  Keeping the proper balance is the trick.  I would love to hear your perspective.

Thanks for the post.  It made me think and then annoyed enough to respond.    

Michael November 15, 2013

Good one!

It all depends, of course. My take on this is that it's not a supermarket size issue, but a learning issue: how does the masterscheduler learn to deal with the specific sort of demand he/she is faced with.

at the gemba, I start with asking this person to keep a single sheet of paper where they write their rules for dealing with demand - we start with the simples case, the few references that make 50% of volume, and we level those.

As they learn we go into increasingly difficult cases, such as seasonal products, as you mention. In this case we will typically level seasonality over the year by treating it as a customer, say, mr russian season, that we produce to stock throughout the year by aiming to have in the beginning of the season the minimum amoun the russian market has ever asked for.

But, of course, the other side of dealing with demand volatility is production flexiblity, which is where the cut these shop stocks in half bit comes in. I am not suggesting to cut finished goods stocks in half, but within-the-flow wip supermarkets, certainly, every time the guys are up to the challenge, because, here again, it's a case of learning.

In this case, the master scheduler progressively updates that rules sheet so it becomes increasingly specific with lost of different ways of handling different cases. The production manager learns to increase change overs (reduce batch size) until the shop floor supermarkets are reduced to the quantity needed by the train every twenty minutes. Both of them learn to work together with obessive care for the pull system (weekly production planning meeting).

This now involves HR (what about people who don't show up for work)? Maintenance (when do we maintain equipment) and procurement (how do we make sure to have the right parts without massive inventories). With each of these functions, we start again the rule sheet and the learning process.

At the end of the day, the first lesson I was taught was lower the water in the lake (the same day the other first lesson I was taught was facilitate operators' work :)). The reason I distrust kanban formulae (not that I don't use them or think we shouldn't) is that they tend to focus minds on optimizing inventory rather than eliminating.

So I stand by "cut it in half", but, obviously, not at the expense of customer service or overburden of the guys on the shop if it's too hard - it's just a challenging starting point to make people think, as it did and we're having this conversation.

How about it?

Phil Coy November 16, 2013

I worked with a company with annual turns of 1.5.  (Yeah, awful and lots of low-hanging fruit.)   They knew they had too much but didn't know how low they could go and didn't know how to know either.   I calculated the EPEI of their processes and told them to change the queue lengths on their SAP MRP to match.  They reduced 30% of inventory in 60 days with no compromise to on-time delivery (and actually no risk to it either since it was based on the math using their historical experience). Their VP Supply Chain confessed to me after that they would have probably chosen to gradually drain the swamp over 15-18 months except they had an extended union shutdown which forced them to do it immediately.

Why do I share this?  Learning is fine and essential for a sustaining the gains but sometimes getting to the math gets a much faster result.  And in situations of high mix/low volume, you can't do it either individually or  manually.
I'm increasingly concerned about the downplaying of the quantitative side of lean which I see all around. 

I'm still chuckling about your approach to seasonality to annualize the impact and build up a season ahead.   Maybe let's engage with some Sales and Operations Planning and come up with a better strategic and cooperative cross-functional approach?

I completely agree that we need to focus on continuous improvement and not focus on "inventory optimization".   Continuing to push for reduced lead time and increased responsiveness will inevitably force lower inventory.  That's a mindset change.   I think it's harder for the materials management types who don't live and die by process to "get it" the same way that the operations types do.  


Michael November 17, 2013

Absolutely agree that we find it difficult to focus on two things at the same time and I'm the first one to bitch about the current dilution of what I call "hardcore" lean - quantitative being a large part of that.

I also dig what you describe, EPEI is definitely where I'd start in most cases. As well as your other comments on cooperation.

Still, in my work, I've found that performance improvement is not sustainable until the guy who has the job has not become autonoimous on the problem we're looking at - in this case until the master scheduler because autonomous in understanding how to smooth-load a high-variability demand.

As per your second point, absolutely, the second sustainability issue I often face is that the quality of the problem solving in the company largely depends on the quality of the teamwork between departments involved.

In this respect quantitative is also key to establishing a joint test method before attempting to solve the probelm across functional borders - that part is always tricky.

Yet, individual problem solving or teamwork seldom occur without executive challenge and patience, and the skill of asking the right questions, hence the "cut it in half" statement. I suspect that the issue we're discussing here, whether with materials management chaps or operations is catchball, when the top dog says 'give me this" and his direct reports come up with the quant objectives and plans of how they intend to do so and they wrangle about this.

This does require training top management at knowing what to ask for :)

Ranjan Choudhary January 9, 2014

We use a solution (interestingly, it is developed on SAP) that calculates the supermarket size for every part on every node of the supply chain, during every planning run.

Besides resizing, it also does several other things like identifying unusual spikes in demand, for which one may choose to create a one time supply and not change the buffer size.

The system also spits out key metrics that go into kaizen activities.