For more than 100 years, Castrol has been a leading name in the development and manufacture of high performance lubricants that keep the engines of cars, trucks and motorcycles running smoothly. The U.K.-based company, now a division of British Petroleum, prides itself on developing products for the most demanding situations-such as Formula 1 racing and attempts at new land-speed records-and then making these competition grade lubricants available to the everyday motorist.
The high performance of the company's products, however, has not always been matched by the performance of its supply chain. In the late '90s, Castrol's European division for commercial and consumer products was struggling with high inventory levels and customer dissatisfaction.
"We had a classical problem," says Alesandro Tenaglia, logistics manager at Castrol Europe, based in Swindon, U.K. "Our customer satisfaction surveys showed very clearly that one of the pains for our customers was low stock availability. They would order 10 items and we would ship only five and then ship the other two, then the other three. This was not good for our customer or for us. Having to make three shipments instead of one and having to make urgent shipments increased our costs a lot. And our inventory levels were also quite high."
At this time Castrol had a minimum of one plant per country in Europe and at least one stocking point per country. The company's customer base varied greatly by region. In some countries, especially Eastern Europe, orders were shipped to a few large distributors, while in other countries orders went to thousands of retail customers, from service stations to mass market stores.
"We tried to analyze our program and understand why we were having these problems," says Tenaglia. "We went with network optimization, trying to reduce the number of warehouses; we tried but didn't succeed in rationalizing the product portfolio. Then we came to understand that the problem was really that we had insufficient or non-existent planning."
Previously, he explains, sales and marketing would make a sales plan based on their best estimate of future demand and then communicate those figures directly to manufacturing, which would create a production plan. "But the quality of the information didn't allow us to make a proper plan for replenishment," says Tenaglia. "It only allowed us to be reactive. So we were in a situation where we had too many warehouses, too many plants and a lot of unnecessary production capacity, but we couldn't reduce those if we first did not do a much better job of planning the operation. What we really needed to do was to improve our understanding of market demand and to improve the communications between sales and marketing, and also our execution of manufacturing, purchasing and distribution."
The company recently had implemented an inventory optimization application from ToolsGroup, Amsterdam, called DPM (formerly, Distribution Planning Model). But Tenaglia knew that technology was only part of the solution. After gaining some experience with the software to understand its capabilities, the European division of Castrol undertook the hard work of organizational change, creating a supply-chain planning department that was totally separate from execution functions.
Tenaglia emphasizes the importance of this move. "Planning was simply non-existent or the function was somehow mixed in with execution, so buyers became planners and manufacturing schedulers became planners," he says. In reality, everyone simply reacted to whatever was happening. "People were very busy anticipating and expediting and they were convinced they were doing a great job, but they were putting all of their efforts in the wrong direction. The point I make is that skills were missing, so we had to first make this big change in the organization.
"Basically, we said to sales and marketing, 'you have been doing the sales forecast, but now we are going to do it using analytical tools that will produce a statistical forecast automatically. Then you can apply your marketing intelligence, but only by exception, to improve the statistical forecast.'"
Forecasting continued at the country level in order to take account of this local market intelligence. Also, Castrol had to accommodate the existence of several different systems among the various countries, each of which used different item codes for Castrol's products. DPM's interface flexibility, however, enabled each country to configure input and output files to fit with the local technology.
One of the top priorities of the new planning organization was to improve the safety stock calculation for each SKU, balancing the trade-off between service levels and the required level of inventory.
As part of this process, the Castrol planning group first decided which items to make to stock and which items to make to order. Then it used the DPM tool to divide make-to-stock items into three classes-A, B and C-with a specific service level attached to each class. "For example, if the Class A service level is going to be 99 percent, this means that, on average, for every 100 orders that I receive for an item belonging to that class, only one order will not find the goods in stock or will have to wait for the stock to be replenished," says Tenaglia. "Obviously, the higher the service level, the more stock you need to support it."
DPM uses unique algorithms to optimize the stock mix for a particular service class by maximizing the service level and minimizing another driver, says Tenaglia. "You can choose whether that other driver is the cost of your stock or the warehouse space or your profitability or whatever."
This is something that simply cannot be done manually, he says. "It helps a lot to segment our items to give a higher service level for those items we believe are strategic and a lower service level for those items where we believe the customer can accept a stock-out," he says. "We know that we will have more stock-outs on some items, normally slow movers, but we try to never have stock-outs on fast movers."
Being able to quantify the cost of service also helps ease decision-making, he notes. Without being able to clearly see the link between cost and service, making decisions about what level of service to provide for different products and customers would involve "a lot of tension," Tenaglia says.
After the plan is created and executed, results must be measured and the plan fine-tuned for continuous improvement, says Tenaglia. This very important third step is another thing that often gets lost when planning and execution are "confused together," he says. "Then, all effort becomes diverted to the actual. If you have an expedite, you arrange that or you try to reschedule production. But you don't go back into the system and say, 'hmm, we need to modify the average delay for this product a little or the minimum lead time for this product.'" As a result, he says, the system's simulation accuracy is reduced. "So the system becomes less reliable and you start relying on people and their personal skills, and that is a kind of groove you never get out of."
Again, he says the answer is to split these two factions clearly and make people on both the planning and execution sides responsible for how well the system predicts what happens in reality.
Another issue facing Castrol was variability in demand caused by promotions. "Basically, demand for lubricants is quite flat, but the sales and marketing organizations in all the countries of Europe push sales through promotions," says Tenaglia. "We have dozens of promotions every year that are very complicated and difficult to translate into impact on volume."
While not part of the initial project, Castrol has since added to its ToolsGroup solution the capability to model specific types of promotions on specific families of products in specific periods of the year. Using unique frequency-based algorithms, the solution plots the "shape" of any given promotion. This same capability improves the forecast by smoothing out lumpiness in demand.
"Our system doesn't just give you a single number forecast," explains Jeff Bodenstab, vice president of marketing at ToolsGroup. "It actually gives you the percentage chance of a series of outcomes-a distribution as opposed to a single number."
Using volatility of order frequency as a key input is what helps smooth demand peaks, says Bodenstab. An example might be a forecast for slow-moving automotive tires. "Let's say once every four months you received an order for four of these tires," says Bodenstab. "The forecast of demand would be one per month, because you are receiving an order for four tires every four months. But what our system does is to actually look at the frequency of the orders, so it will tell you that you need to stock up to accommodate an order that is much larger than the actual forecasted demand-in this case four times as large. Now that might sound really simple for someone who happens to know that tires are usually ordered in quantities of four, but what happens in the real world is that you have tens of thousands or even millions of SKUS with lots of variations-some items are ordered in small quantities, some are ordered in large quantities-and you really can't track all of that manually. But if you understand that lumpiness, you can set your inventory targets to better accommodate it, so you get higher service levels without actually increasing your inventories."
Tenaglia says Castrol is "still struggling to improve our ability to predict promotions.
"To be honest, the big impact we got on demand planning was introducing a statistical tool that enabled the translation of our sales forecast at an SKU level for the next 18 months, with no manual work. This freed people's time to do value-added work."
The initial phase of the project generated significant inventory savings as well, though Tenaglia acknowledges that "the starting point was bad. We were able to reduce by 20 percent our inventories in the first year and another 20 percent in the second year, which is about a 38 percent total reduction," he says. Surprisingly, most of these reductions were in raw materials, which at Castrol includes base oil, additive, labels and packaging. "Once we were able to smooth the production plan and increase the reliability of future production plans, the necessity for these raw materials disappeared, so we had a reduction there of 9 percentage points," he says.
One key to achieving these results was building in-house expertise. "We decided not to ask the software company to recommend the best implementation or to configure the system," says Tenaglia. "Instead, we created a group of people to build in-house knowledge. It took a long time and had high costs, but at the end we had a core group of people with deep expertise who have been able to configure the tool and make it work best for our operations. I think the decision to go with in-house knowledge building was a key success factor."
After a couple of years, however, people started to change jobs or move out of the company and Castrol began to lose that carefully built knowledge. "One of the lessons we learned is that you not only need to build the knowledge and the skills internally, but at the same time it is necessary to create a continuous training program so that the people who become expert can also become teachers," says Tenaglia. "This is something we are trying to implement now, because without the expert people to mange the processes it is getting more and more difficult to grow and improve. This means that we need to change our rewarding policy and so on."
Personnel changes were not the only bump in the road for this project. A major hiccup occurred in July 2000 with the acquisition of Castrol by BP.
"BP was at the stage we were before we started this project," says Tenaglia, "but we were not in a position to impose operational principles because we had been acquired. So we lost another couple of years before the BP people understood that we definitely needed to get excellent in sales and operations planning."
With that realization came the project's second wave, which included an expansion and rollout of the ToolsGroup solution, which now spans 29 installations in 25 countries in Europe and South America.
The second wave also brought enterprise targets for increased service levels, which created a new challenge. After the acquisition, Tenaglia explains, Castrol Europe had 20 different country operations and 10 different ERP systems. With that many different systems, he says, "when you ask for a KPI, even if the definition is very tight and precise-which ours was not-it is very difficult to really get the same numbers. There is a lot of room for interpretation."
This was demonstrated when every country division reported hitting the new service-level targets. "At a certain point in time it appeared that 99 percent of our orders were in full and on time, which means that all order lines have been served in time and in full for the quantity. This was simply impossible. We could go to any country and ask to see a stock-out list and it would be a couple of pages long."
As a further test, Tenaglia sent out an exercise to eight countries. He gave the managers raw data of the forecast and actual demand for products and asked them to calculate the forecast accuracy using a global formula. "We got eight different results," he says, "and the formula is very, very easy. Everyone was doing something a little different. They would round numbers or they believed these items should be excluded or this business should be separated and so on. This clearly is a reflection of the need to improve process performance."
Castrol is addressing this issue in two ways. First, it is collaborating with its business units on developing better definitions for KPIs. And it is using ToolsGroup as the standard for calculating the KPI. "If you don't only have the same formula but the same technology, then you can guarantee that the result in consistent," says Tenaglia.
The company also is working on a new set of KPIs that better reflect process performance. "One very simple KPI that we have today is in full, in stock availability, but this doesn't really reflect process performance," says Tenaglia. For example, he says forecast accuracy is determined by three factors: demand variability, which the company doesn't control; the ability of marketing intelligence to improve the statistical forecast by better understanding promotions, price changes or competitive activity; and product complexity. "We have tried to calculate a KPI for each of these variables," says Tenaglia. "In particular, we do not want to measure people against forecast accuracy but against their ability to improve the statistical forecast."
It is important to be very careful with this, he says, because every time a KPI is part of an employee's performance contract, it can introduce a distortion factor. "Of course, people look at their performance contract rather than the company's performance," he says.
Another factor facing Castrol is BP's decision to standardize its ERP on a single SAP implementation. While Tenaglia expects this will mean a move to APO for supply-chain management, he does not believe this will affect the company's use of ToolsGroup. "I expect that we will keep DPM as a bolt-on system," he says. "DPM is a very powerful approach with a planning capability that is unique in my experience."
Timely, incisive articles delivered directly to your inbox.