Imagine the scenario. You need to introduce a new product to market. But building a dedicated production line for that product will cost you around $500 million. On the other hand, if you were to invest in software that allowed your plant to schedule this new product along your existing production lines – and it would cost you a mere $3 million to implement – which path would you opt for? Two years ago, Nissan’s Sunderland plant was faced with this very same issue, and chose the latter option (APS: advanced scheduling software), and it didn’t have to move mountains to see business benefit. As Michael Simpson, production controller and project leader, points out: “Since the [APS] implementation, we’ve seen a 30% improvement in plant productivity, and we haven’t had to invest huge sums of money on a third, completely new, production line. We simply schedule orders for all three cars [Micra, Primera and the new Almera] down the two existing ones. We’ve saved millions!” Nissan’s site is one of the best: it’s renowned for its productivity (currently ranked 10th most efficient plant in the world), with a ratio of 94 cars manufactured per direct employee per year. With 4,700 employees, it produces around 337,000 units each year, and exports 75%. It’s the largest volume UK car plant, and has more than 200 European suppliers, some on-site. The scheduling issue arose in 1998 when Nissan decided to introduce the Almera model to Europe. This meant Nissan had to look for extra manufacturing capacity from one of its global sites. It eventually chose Sunderland, but was concerned the plant should not be disrupted from its existing productivity level. At this time, Sunderland had two dedicated assembly lines, one for the Micra, one for the Primera. But Simpson adds: “Installing another dedicated assembly line for the new Almera would have taken a year to complete - time the company simply didn’t have - and an investment of $500 million.” Bottleneck problems The production process was typical for the industry. There are three ‘shops’: body shop (where the initial chassis is assembled), paint shop, and trim shop. Between these shops was a buffer storage area where autos that finished the previous step would wait to be processed by the next one – the idea being that autos spent as little time as possible waiting in these areas. But as Simpson points out: “Assembly lines of this type present a very significant scheduling challenge. For example, changing the colour of the paint in the paint bays is very time-consuming. So, as much as possible, cars of the same colour need to be built in the same batch. And sunroofs add several additional steps to the assembly process, so these cars cannot be scheduled back-to-back for fear of holding up the entire line.” Simpson saw these as the system constraints. The body build facility scheduled production on a shop-by-shop basis. So, the derived schedule reflected each shop’s constraints and was co-ordinated from a control room. A problem in the process resulted in meetings between shop managers who would elect some ad hoc solution to the sequencing issue, which would then be managed by the control room. “Using this approach,” explains Simpson, “as much as 90% of the cars would require some sort of re-sequencing, adjustment or control room management. We used two people, full-time, to re-sequence vehicles through the line. The home-grown mainframe system [which produced a flat text file order schedule] simply didn’t create a good enough sequence.” With two cars being produced, the system scheduled one week’s orders, with one model only per production line. “Manual amendments to this schedule were taking about two man weeks in total,” explains Simpson. “And a mere 3% of vehicles actually passed through the line process and came out the other end in the correct, planned sequence.” Clearly something had to change. So Nissan invited tenders from prospective suppliers including Ilog, i2, and Lanner Group. After a two-month evaluation, the decision was made to go with PA Consulting with Ilog as software supplier. “It was an intensive evaluation period for us, and we chose Ilog and PA partly because they were prepared to do on-site evaluation, whereas certain other vendors wanted to do it their way,” explains Simpson. “Price was not really an issue, although one of the vendors was disregarded for being far too cheap – they clearly didn’t understand what we needed from them!” So, in March 1998, a project team of 24 was assembled: 12 PA consultants and 12 Nissan staff. The team tackled the problem systematically: first, they had to consider how cars were being scheduled at the plant, then determine how a third, virtual line could be intertwined with the two existing ones. The solution was end-to-end scheduling through the complete body build facility, using Ilog Solver (APS) software. The system is an optimisation package based on constraint (rather than linear) programming. This means users can specify the constraints of a process (such as parts availability, painting restrictions) and the software then generates a production schedule for the orders that need processing. Using Ilog Solver, the team developed a customised solution whereby the new Almera was constructed solely on Assembly Line 2, and the previous Micra was now assembled on both assembly lines since it was the simplest car to manufacture. There are now multiple, physical cross-over points (eg: overhead conveyors) between the two lines and 16 possible paths the Micra can take during assembly. “Which path it [the Micra] takes depends on what makes most sense on any given day,” says Simpson. Immediate ROI The software was designed and tested by the 24-man team over 12 months to August 1999: one month fact-finding, identifying constraints (there were 2,500); two months specification work; seven months for design and build; and the rest was user testing and system release. The project was deployed on time (ready for the January 2000 launch of the Almera) and on budget. The ROI was immediate, as Nissan was able to produce a third model with very little capital investment – around $3 million. Simpson adds: “Capacity has increased by an eye-popping 30%, and only 5% of cars now require intervention from the control room! This is testimony to how well it all works now.” And it is impressive. Customer orders (one week’s at present, soon to be daily) are now imported from the mainframe system into Ilog, which then generates a vehicle sequence (in minutes) for the forthcoming week. Solver intelligently considers all the constraints and build rules for each batch of orders in arriving at a finished sequence. Simpson explains further: “It now takes only four hours to generate a schedule, and we can run different test scenarios to try to optimise the batch sequence.” There are three Ilog users in production planning and control. Solver sits underneath an Access database on standard desktop PCs. Ilog Views sits over the top of everything, providing a very user-friendly graphical user interface. The PCs are connected to Nissan’s existing LAN (local area network), so no upgrading was necessary. Only a new server was required, but this was part of the implementation cost. “The results speak for themselves. There are no manual amendments to the schedule. We have a 95% straight through ratio [ie: in the original, planned sequence] and operators know exactly which assembly has to be fitted next. It’s all about ensuring the right parts are available to the right people at the right time. That’s what APS is good at. “And a better build to schedule ratio means we can operate with less stock and a more synchronous supply of parts to the assembly lines. We now send Access reports of the planned schedule to suppliers to help them plan more effectively.”