Smooth is smart
Informed opinion states that, sometime in 2006, Toyota will become the largest automaker in the world. So, this is an appropriate time to examine the hidden virtue of the Toyota Production System (TPS), which is the engine behind the growth of the company.
Apparently well-documented in both Japanese and English, TPS is much emulated throughout the Western and Eastern hemispheres. Despite this frenzy of activity, research shows that the overall success rate is relatively low. Implementation of TPS frequently fails, even when a company’s suppliers are advised by Toyota consultants. It helps to understand TPS theory. Smooth is smart (SIS) is the hidden virtue behind TPS. Reducing uncertainties from all sources (see the ‘vicious uncertainty circle’, fig 1, p30) is the key enabler in successful implementation.
To achieve this goal it is an essential first step in applying TPS concepts to put one’s own house in order: to have reliable and consistent processes that respond exactly as required. It is here that lean production has a major part to play; since non-value added time – the reviled waste activities – are removed. A major effect of this activity is the beneficial reduction of Work-In-Process (WIP) and indeed this has been substantially reduced in many, if not most businesses. If this internal waste removal programme is executed properly, there will be other significant benefits accruing to the company. On top of the ability to deliver products as and when required at competitive cost, the business will have acquired considerable skills in change management and waste elimination. In this way it will be in a position of leadership (via knowledge internally acquired and applied) to implement supplier development programmes to remove waste from the upstream part of the supply chain. Such opportunities for transforming the vendor base into an efficient and effective pipeline means that this level of enterprise has become very competitive (fig 1). But again there are further benefits. By working closely with suppliers of our present needs we are moving towards future liaisons with steeper learning curves (hence being more responsive) and lowering transaction costs generally.
Production smoothing Japanese style
To determine the extent of real-world production smoothing, economist André Mollick has collated published performance figures available for the Japanese Automotive Sector over the from 1985 to 1994. Such data is available monthly, in this instance yielding a reasonable time series for statistical analysis. Nine categories of vehicles are studied, including cars, buses, trucks, motorcycles, and cycles. The explicit purpose of the model summarised in table 1 (p32) is to demonstrate beyond reasonable doubt that the Japanese vehicle industry actually smoothes sales orders when setting production schedules. Yet the inventory-to-sales ratio expressed as months of stock is as low as 0.39 (for light trucks), and even in the worst case is no greater than 1.14 (for large buses). Hence annual stock turns are reasonably impressive, averaging out at around 20:1. So the Japanese vehicle industry has achieved the enviable distinction of making-to-stock, but coupling this with minimum reasonable inventory (MRI), as distinct from zero stock.
What does the statistical analysis in table 1 actually mean? Using a robust methodology, Mollick has tested the published data via three criteria of particular interest to operations management. The model thereby demonstrates the existence of make-to-stock policies; links between sales and stock levels; and evidence on significant on-costs due to production fluctuations. The latter are of particular interest since executives often fail to recognise just how much it really costs to keep ramping production levels up and down. To make comparisons easy, the three factors studied have been highlighted in table 1 for those cases where the evidence is considered to be highly significant. Hence we can be reasonably confident that these effects are unlikely to be due to pure chance.
Pass the good news up the chain
Before moving on it is worthwhile examining table 1 in more detail. Simple counting of high correlates results in suggesting that strong evidence exists on make-to-stock policies in (7/9) product groups; on links between sales and stocks in (8/9) product groups; and there are (6/9) product groups with significant production on-costs incurred if they move away from a level schedule. Perhaps it is a more powerful pointer to suggest that in (21/27) cases listed in table 1 there is a high level of proof, or even to note that each product grouping qualifies on at least (2/3) of these criteria. The important outcome is the realisation that there is a high likelihood that the Japanese Vehicle Industry has apparently reached the holy grail of smooth production coupled with low stocks. Obviously it is now time to move on from the what to the how. The detailed description of the Toyota Production System by K Suzaki (1987) with his emphasis on level scheduling is a good starting point for the uninitiated.
It is essential that the production of different items be distributed evenly so as to minimise uncertainty for upstream operations and suppliers. Toyota defines level production as evenly distributing volume and specifications over the span of production during the assembly process so that suppliers have a smooth stable demand stream. A non-level schedule (e.g. making a large batch of two-door cars and then switching to a large batch of four-door cars) presents suppliers with spikes in demand. To deal with these spikes, suppliers are then forced to maintain high inventory levels to ensure they are prepared for unexpected customer demand. According to JK Liker and YC Wu (2000) suppliers often call the finished goods inventory maintained for its non-level scheduled customers ‘the customers wall of shame’. As we shall see, there are a number of prominent OEMs that qualify for this description. This is primarily due to the ‘front-end’ uncertainty induced within the vicious circle of fig 1.
The TPS view is that non-level customers are simply imposing their chaotic requirements onto their suppliers. Providing suppliers with a predicable schedule, stable orders and regular demand is particularly important in a logistics system with little inventory to cover unexpected fluctuations in demand. The data illustrates Japanese transplants readily outperform their US counterparts on this dimension. Table 2 shows their order demand is stable, reflecting effective level schedules in their assembly plants. Honda, Nissan and Toyota stand out among their Japanese peers. Toyota is so sensitive to the importance of level production that the automaker states those customers unable to level their schedule do not have the right to use a JIT pull system (which experience shows can suffer from lack of robustness if so abused).
Strategic use of waste
Despite advertising the availability of a wide range of variants, customer feedback, trade reviews, and expert reports dictate that only a limited number of models will actually be produced in reasonable volumes. Only these will have a short delivery time quoted. They will apparently be made to order, but in fact as we have seen are made to limited stock (MRI). It is the ability of our customer – in this case the auto-retailer – to influence final purchaser decision, which is a big factor in SIS. The close working of Japanese style automaker and the customer enables SIS to function. The retailer provides the OEM with rapid feedback on sales trends, including end customer detailed choices. But the OEM updates the retailer with information on which variants are likely to be available in the near future. Hence the retailer subtly influences end customer thinking into ordering a vehicle likely to be available ‘just-in-time’.
How does Toyota then pinpoint which retailers need further working on to achieve the necessary reduction in demand uncertainty in the vicious circle? Management resources are an expensive and scarce item, so focus is essential. To reach this goal, TPS abandons the religious theme of waste reduction temporarily. In the process it moves away from conventional western wisdom on the pooling of inventories. Instead the Finished Vehicle Stocks for each retailer are compartmentalised separately and displayed transparently on wallboards. Such charts are all that is needed to identify any retailers imposing volatile demand on the delivery system. So TPS tolerates strategic waste in the short term in order to facilitate level scheduling in the longer term.
What of that awkward customer who insists on ordering a peculiar variant? In fact carefully adjusted variable delivery times to the end purchaser are not just the prerogative of TPS. The old UK Lucas Group was well-known for categorising products into runners, repeaters, strangers, and aliens, and prioritising production on that basis. Lucas, it must be remembered, were classified by no less a dignitary than Richard Schonberger (1996) as world-class first-tier suppliers. Their performance was reckoned by him, on the basis of published data, as being every bit as good as their Japanese competitors.
The well-trodden path to success
Paul Childerhouse and Denis Towill published substantial industrial evidence on the best way to proceed to engineer TPS from the traditional supply chain. This well-trodden pathway is shown in fig 2, and is backed up with statistical analysis of audit results from a sample of value streams ranging from baseline to minimal uncertainty operations. To understand the preferred route to improvement, value streams are regarded as a system in which there is our own delivery process, preceded by the supplier side consisting of all our vendors. Downstream is the demand side driven by our customers. The controls side is operated by our business based on the current information and any forecasts that may be available. Hence the controls side is responsible for placing internal orders for our output to best balance supply and demand.
What our investigative results clearly show is that the vital first step is to improve our process. We therefore eliminate scrap, minimise total cycle time and minimise set-up times. This ensures that what is ordered is delivered exactly as and when required. It is the knowledge gained during this ‘lean production’ phase that can then be exploited via supplier enhancement and improved vendor interface management. Now our business capability is greatly strengthened and can offer the customer effective and reliable delivery. Clearly the controls side can be improved sequentially after each phase is executed and in the light of the new and more reliable data received.
Unfortunately, there is still the downstream element of our original vicious circle to contend with. But by this time our business is well up the learning curve. So we are in a position to negotiate for greatly improved customer interface management. This includes both sales and inventory transparency. The final stage is to work so closely with the end customer that what is sold and what is available are in unison. With the market strength of Toyota this manifestly results in consistent volumes and low volatility. Variety is no longer a surprise either, since variable delivery dates can be quoted as necessary in retailer negotiations with the purchaser.
Collaborative virtuous circles
It may be argued that rocket science may be used to prove that advance information is a good thing to have. If so, a conclusion has been reached that may be regarded by Toyota as intuitively obvious. However, to players in the system under heavy pressure to deliver across a wide range of SKUs (stock keeping units); more information can just lead to more confusion. Hence the need for the available information to be noiseless, unbiased, and clearly and simply shown within the systems context. Making level scheduling (the simplest of all control policies) work with each retailer means TPS smoothes production in a virtuous circle as shown in fig 3. But working closely with our customer has the further benefit of enabling us to avoid costly and time consuming knee jerk reactions to changes in capacity requirements as well as predicting production levels reasonably accurately. A hidden secret of TPS, as we have already seen, is passing this good news up the chain so that all our suppliers benefit from level scheduling. And if this is not possible, then we smooth as much as we can. Single minute exchange of dies (SMED) does not mean we are willing to change the product every few minutes.
*Please note: This article first appeared in the April 2006 edition of Manufacturing Engineer magazine.