Lean is based on the thought that excess work (muda) needs to be removed. So we can lead this into the variation removing. We must still remember that variation can also be good if it’s our strategic decision.
This post is part of bigger series about variability and WIP and I recommend you to read these posts first:
– Is it crucial to control WIP
– How does the variability affect on production
– Variation in Practice – Part 1
Example Ford was successful because it simplified car manufacturing into one model which enabled decrease in production costs and more people had the money to buy car. But in 1930-1940 General Motors (GM) started to offer more models and won some market shares from Ford which nearly drove Ford to bankruptcy.
So the important thing is to ponder what is your strategy. Variations should be able to price so that they cover the costs they are causing in production. In other words car manufacturers are in business to make money not in the business of removing waste (muda). Money can be done only by satisfying customer needs better than competitors.
Example some oil change shops offer oil change without booking. They have to face some variation but if they start to book oil changes they will ruin their whole strategy and competitiveness.
Variation is always buffered with inventory, capacity or time. Or some combination of these. Example office supplier need to stock enough pens. If they are sold out no one will wait their ordering. Customer will walk somewhere else to buy it. So the demand is buffered with inventory. In all products that are needed immediately (which this cheap pen can be included) demand variation is buffered with inventory. Because product is needed immediately it can’t be produced which means that extra capacity won’t be an alternative.
Another example is rescue services. Happening of fire or stroke is both very random and rare. Events are highly variable and they can not be scheduled or predicted. Because these events need urgent respond we can’t buffer them with time. We can’t also stock ambulance trips. So the only way to buffer is capacity. We need to have enough capacity in order to handle these events. In practice the utilization of fire trucks is very low.
Demand for organ transplants is highly variable such is their supply too. Because we can’t stock organ transplants and we can’t ethically increase their supply (increase deaths) the only way to buffer is time. Indeed the queue for organ transplants is long.
These examples are great illustration of how variation is always buffered with inventory, capacity or (waiting) time. Think about your business how things are now. Would it example be reasonable to move buffer from inventory to waiting time?
TPS – Toyota Production System
So how did Toyota manage variability?
Demand variability. Design and marketing was so successful that demand continually exceeded supply. This helped in many ways, like possibility to narrow models. Also many of the add-on parts were assembled by dealer*. And because of big demand the production was scheduled months ahead. All this practically removed all variation from productions perspective.
Production variability. By focusing on reduction in setup times, work standardization, quality, error proofing, preventive maintenance and other flow improvements Toyota was able to significantly decrease production variability.
Suppliers variability. In the early 1980s Toyota had good negotiation power for its suppliers. Because Toyota produced big part of their revenue Toyota had good leverage on them. Toyota’s directors even sat on the boards of some suppliers. This ensured that Toyota had its parts whenever they wanted, suppliers adopted some work methods that Toyota proposed for variability controlling and suppliers took care of excess inventories.
Also the rest of the variability in production was taken care with capacity buffer. Shifts was scheduled less than 3 per day and end of the shifts was used for preventive maintenance. With these and top class waste-less production the daily throughput was highly predictable.
* In one theoretical case we designed motorbike production where plastic parts were added to shipping container in warehouse. This saved time from production and made it easy to sell different colors for customers. Assembly was done by dealer or customer. We also decided to ship multiple variations of tapes. Cost of one set of tapes is nonexistent compared to the time spent taping them on production line and variation caused by different tapes. It is much much cheaper to ship multiple set of tapes to customers and let them decide what kind of tapes to put on their bikes if none.
Cost of variation
Variation can be playfully called “pay me now or pay me later”. This means that if variability is not controlled we have to pay from it with lost throughput, capacity, long production times, large inventories or bad customer service.
Lets take an example of production line with two machines. Production time for bot mahines is 20 min and their CV is 1, which is medium variation. One job includes 50 parts and between the machines there is WIP storage for 10 jobs (500 parts).
Optimal WIP would obviously be 50 x 2, one job per machine, total WIP of 100 parts. But because of variability this can not be achieved.
|Case||Buffer (jobs)||Process time on machine 2||CV||TH (/day)||Utilization||CT (min)||WIP|
|1||10||20 min||1||3321||92,3 %||150||347|
|2||1||20 min||1||2712||75,3 %||60||113|
|3||1||10 min||1||3367||93,5 %||36||83|
|4||1||20 min||0,25||3443||95,6 %||51||123|
Table above illustrates the effects of different options. First case is when we use our full WIP storage of 10 jobs. We will have output of 3321 parts per day instead of theoretical maximum 3600 parts. This throughput is achieved in cost of big WIP and long cycle time.
In second option the medium storage is decreased to 1 job when our throughput is dropped to 2712 products per day. Without WIP storage the second machine will have to continually wait for parts. This decreases the utilization of second machine so the lower WIP is caused with the cost of lost capacity.
In third option our second machine is replaced with faster machine which CV is still 1. In this option cycle time is decreased to 36 minutes and WIP is only 83 parts. Overall TH is little better than in first option. This is achieved with the cost of lost capacity because the capacity on second machine is utilized only under 50 %. That’s why this is not reasonable option if machine is expensive. If machine is cheap this can be quite reasonable and easy way.
In fourth option we have been able to decrease the variability from 1 to 0,25. Other variables are same than in second option. This leads us to 95,6 % TH, 3443 parts per day, WIP level only 123 parts and CT of 51 minutes. If we are able to decrease variability this is the best option.
Buffer required by variability can be somehow moderated with flexibility. Example the safety stock example in previous post the stock was made more flexible by stocking components for assembly instead of stocking pre-assembled computers. Correspondingly many computer part suppliers produce only the main product and the country specific power cable is put in the box in dispatch.
Flexible workforce that can do multiple tasks can do work there where it is most needed. This way we can cope with lower workforce than if workers could manage only one job.