Customer pulls the strings

'We don't manufacture anything unless there's a customer demand'. How many times have you heard that and thought 'yeah, right', because getting manufacturing properly demand-driven is a very difficult thing to do? Lean thinking certainly helps; Theory of Constraints, less popular today, also sheds invaluable light on the notion of bottlenecks; and there are plenty of other techniques. But the plain fact is that, when we attempt to sequence whole factory floors or, worse, complete businesses, to the beat of the customer demand drum, practice turns out to be more complicated than theory. We all know why: too much demand variability (in the worst case, outright seasonality); too many product options; and shared resources necessitated by factory size and simple economics. Add to those: variable component and assembly lead times; convoluted supply chains; unpredictable plant breakdowns; quality and/or yield problems; and late changes to customer orders. The list goes on. So it's always fascinating to find a manufacturer still claiming, 'Yes, we really don't manufacture anything unless there's a customer demand'. Texecom, which manufactures electronic alarms and security systems is one such, and operations director Russell Trotter concedes that getting there has only been possible by dint of an aggressive change programme and rethinking the company's IT, which to date has taken more than two years. "But now, from a purchase order being raised all the way to procurement needing parts from the supply chain, everything is pulled from the initial customer order," he insists. So let's explore this phenomenal claim. "We looked at three main value streams – procurement, logistics and manufacturing control. I'm not talking about traditional process streams, but data streams, and we treated them as business processes, stripping out non value-add," explains Trotter. "Apart from that, the other main strategy was driving down stock holdings of parts and finished goods throughout the facility… We now have just thee quarters of one week of finished goods [because that's the first to fulfil customer orders and start the pull process] and virtually zero WIP, with absolutely nothing part-built. All we have is one PCB buffer position [because output is limited by how fast the electronics line can build] and that's it." 'Aha,' I hear you say, 'so there is some material manufactured ahead of demand.' True, but only just. And give the guy a break: this factory and its underpinning supply network is one of the true exemplars, upon the shoulders of which we might attempt our own transformations. If we follow some of Texecom's approaches, the rewards may well be significant and sustainable cost savings alongside excellence in customer service. For a little background, Trotter explains that, from a procurement point of view, Texecom has 650 base builds with 1,800 different purchase parts, coming from 70 suppliers. It also consumes 200,000 PCBs populated with 28 million components every month. Beyond all that, the company runs 24 plastic injection moulding machines and the picture is completed by 35 final assembly cells, with 2,400 stock locations across the business. These account for 200,000 stock movements per month resulting in 2,500 despatched orders. "On procurement we were manually raising 1,500 orders per month covering 5,000 line items," says Trotter. "Now, we're down to 5% still ordered manually, with all the rest automated. We did that by using the MRP module in our Epicor Vantage ERP system, but treating it only as a data management and reporting tool, with everything effectively on electronic kanban triggers. Epicor runs a stock control check every evening and automatically signals our suppliers, via EDI, with the requirements for the next day or two, according to their lead times." What about suppliers that can't cope with EDI? "For them, we've implemented virtual consignment stock – 'virtual', because we don't like the space it consumes, so we just manage it over the web on their premises." Backflushed component usage acts as their signal for replenishment and they invoice Texecom against that, too. "So we've stripped the waste right out of that value stream and, now, our purchasing manager can focus on managing – ensuring, for example, that the cost of components is maintained." Trotter isn't suggesting that his suppliers are all capable of turning on a sixpence. "We set up a five month forecast and it's up to them to gear up to fulfil," he explains. And making that work in the short term is about a monthly review of the historical trends for its 650 base builds, with special offers and planned obsolescence factored in – then running MRP to simulate demand and issuing updates to suppliers. "Those are pretty much firm, plus or minus a few percentage points, and it's the suppliers' job to make it work," he says. And here's the contrast: "Four years ago, MRP ran and created a demand on manufacturing to build, using batch operations. We don't recognise those words any more. There are no works orders onto the shopfloor – just automatic reports that tell supervisors they need 'x' amount of 'y' products in fished goods. Those ripple back through the cells and on to the electronic PCB buffer position – always with priorities determined automatically by kanban percentage fill rates – and so on out to the supply chain. That's one reason why we don't need planning and scheduling software." Now there's slick. And it doesn't stop there. At a logical level, Texecom uses its ERP in another slightly different way, to help it handle events and change. "Our data includes processing times for every product for each of the cells, so once a month we're able to run a 'hot spot' report that tells us the loading for each cell, and whether it's getting close to capacity. Also, every day we get advance warning if the order input rate looks like it might cause concerns: we can see the percentages trending towards critical levels long before the point at which the buffers would be in danger of not recovering." It's a similar picture with logistics, although Trotter admits this is still work in progress. "There are two issues here: making sure parts arrive at lineside when necessary and making sure there are no mistakes, in terms of supplying the wrong components – especially to the electronics production line," he explains. "Making that work has been about getting our stock locations identified with barcodes, and we've still got a little way to go. Currently, stock movements are managed by the logistics team scanning barcodes and stock holding positions, but ultimately, we'll get away from that. Each cell will use the same system but backflush used parts and put demand into logistics that way." Beyond that, there are what amount to barcode-based interlocks on Texecom's surface mount electronics machines to prevent expensive errors. "And the other thing is that all finished goods moving out of the manufacturing facility are serialised to the individual order, with their own barcode. That has streamlined distribution, because operatives no longer have to think about what they're picking. They just scan the kit and it's onto the pallet. But also, it has completely eradicated parts 'going missing' in transit. Our product is highly desirable so there always used to be a certain amount of wastage in the supply chain. Now that's history." And finally we come to manufacturing control, which Trotter describes as pulling everything together. "When we first started looking at it, the complexity of production meant people were putting a lot of time into attempting to optimise what to make that day. They were changing this, changing that, stopping moulding operations of this and changing it to that. So we reorganised, as per the logistics and procurement streams – in this case, changing manufacturing control to focus on whether the overall system had the correct buffers." Getting people on side, he says, was achieved by giving them a say in what the buffers should look like. "We started by establishing how long it takes to manufacture 'x'. Then, if it takes 'y' minutes per item and you need 300 per day to deliver in a two-day window, you can see the buffer you need in fished goods and in electronics behind that. Initially, we exaggerated the figures to allow a safety margin, but, since then, we've been chipping away." Interestingly, he indicates that this process continues for new product introduction. "If it involves an existing part base, a new product is a non event for us. However, if it's not a part we've previously identified in the forecast to suppliers, then we start with the suppliers' lead times before pulling back to the same as everything else. In the meantime, we talk with marketing and sales and build, say, 400 for finished goods, and then watch the sales history to get that buffer right. The rest works itself out." For Trotter, though, the final big deal here was education. "The task was to wean people off the way they had been working before – by proving that, as long as they followed the new methods, maintained data accuracy and watched the buffers, they didn't have to worry about which items should be built in what order." Summing up, Trotter agrees that the thinking throughout is essentially around logistics planning and control, as opposed to production planning and control – which now manages itself. But, for him, there is one more key point. "We're in the very fortunate situation that we have a dedicated systems development resource. In most companies that resource simply doesn't exist. Manufacturers may have people who know about ERP in their own areas but, if they want to get into report writing, SQL etc, and make the kinds of changes we've made, they're into external consultants and big money. "What I would take with me is the importance of getting a dedicated IT resource that absolutely understands the whole business. All the other work is great, but, if you can't get an expert on the IT side you will struggle." Additional DDM IT strategies For companies whose manufacturing resembles project-based or engineer-to-order operations – or whose product portfolio, lead times and order mix are too variable and complicated for ERP-assisted pull signalling alone – APS (advanced planning and scheduling) software provides the essential sequencing and 'what if' functionality. EADS subsidiary Eurocopter, for example, chose Preactor to make sense of its Dauphin multi-role helicopter production line, cutting the assembly cycle from 30 to 16 weeks. Running on top of SAP, Preactor now helps planners to optimise shopfloor sequencing of complex assembly work, managing 45,000 parts, with 800 operations and WIP for 25 aircraft – recognising real-world supply chain issues and enabling otherwise impossible lean pull operations. The many advocates of sales and operations planning will be interested to learn that ERP giant Infor has released Infor S&OP v1.0 specifically to assist with demand-driven operations. Bruce Richardson, chief strategy officer at Infor, explains that it's all about quickly, accurately but also profitably adjusting operations against demand. "Any solution must align manufacturing, sales, finance and operations around the most profitable and achievable build plan," he says – adding that the new generation software will provide "unprecedented capabilities for creating, managing and using electronic workflows that address every major stage in the S&OP process". 'Supply chain as a service' is, as yet, an uncoined phrase – but it's a good way of thinking about GXS's Internet-based Trading Grid and its associated outsourcing capabilities that currently look after supply chain functions for 75% of the world's Fortune 500. GXS has huge pedigree in running web-based supply networks, supporting all the EDI variants, providing portal coverage, managing electronic processes and documentation etc at one level, but also sourcing and 'on-boarding' appropriate suppliers, whatever their level of IT competence.