A decade ago the Household and Body Care Group at Sara Lee Corp. instituted a formal Sales and Operations Planning (S&OP) process to better align supply and demand. Key performance measures improved - up to a point - then hit a stone wall. The culprit: unyielding inaccuracies in the sales forecast. It was only when management took forecasting away from marketing and gave it to professional demand planners equipped with new software tools that the S&OP process really began to hum.

Chicago-based Sara Lee is better known for desserts than for household and body-care products. But with brands like Kiwi in shoe care, Endust and Behold in furniture care and Ambi-Pur in air fresheners, the household division is an important contributor to profitability. For the quarter ended Dec. 28, 2002, for example, Sara Lee said volume gains in household products and meats offset volume declines in baking and beverage, helping the company more than double its net income compared with the same period a year ago.

That improvement also reflected a winding down of restructuring charges over the past two years, during which Sara Lee focused more resources on increasing sales in key brands. Focusing on brands is part of an overall de-verticalization strategy, under which the company earlier shed its manufacturing assets - an unusual move for a consumer-goods manufacturer. "The company wanted to focus its assets and brainpower on building brands, not on how to make the best widgets," says Gary Kahler, director of S&OP at Sara Lee Household and Body Care. "It was a strategic and a financial decision."

Manufacturing divestiture reached the H&BC division about three years ago, Kahler says. "We sold off our manufacturing plants at that time and became dependent on third-party supply." The largest plant operation, which supplied shoe-care products that represented about half of the division's cost of goods sold, was located only 15 miles from its base outside Philadelphia. Hauer Custom Manufacturing purchased that facility. "We immediately entered into a supply agreement with Hauer and it still makes a high percentage of our shoe-care products," Kahler says.

The supply chain for other products was lengthening, however. Already the most global of Sara Lee's business units, the household division was working to more closely align brands in all its regions, particularly North America and Europe. As a result, the North American business was importing more products. Ambi-Pur, for example, was a successful European brand before being introduced in the U.S. and Canada and continues to be sourced from there.

"We have an increasing volume coming in from Europe and the lead time for those items can be 12 or 13 weeks," says Kahler. "Obviously, we have to carry more stock to cover that longer lead time."

These changes only exacerbated the division's problems with forecast accuracy. "The forecast variances we were experiencing tended to be fairly large," says Kahler. "That required an inordinate amount of safety stock, and we still sometimes wound up having stock-outs."

At that time, three years ago, marketing was responsible for developing sales forecasts as well as associated financial forecasts. "By their own admission, marketing devoted very little time and attention to the forecast," says Kahler. "What they wanted to do was market. They weren't the least bit interested in forecasting and they didn't do a very good job."

For that reason, the company decided to create a demand-planning group. Kahler was given the OK to hire two demand planners, in addition to the existing forecast manager, who had been coordinating the forecast with marketing. Responsibility for forecasting and demand planning was vested in this new unit, under Kahler's direction. "Getting the people resources was important, but I also needed better tools to do the job," he says. "We had an old DOS-based forecast package that we acquired in the early 1990s from Lucas-Bear. We wanted to upgrade that to a Windows platform and get the 10 years of additional sophistication that had been built into demand planning software during that time."

Narrowing the Field
Kahler and his team looked at several options, including the latest offering from Lucas-Bear, but quickly narrowed the choices down to Prescient, West Chester, Pa.; Mercia Software, Atlanta, which Kahler knew from work at a previous company; and the demand planning modules that came with the enterprise resource planning (ERP) software of SAP, Newtown Square, Pa., and J.D. Edwards, Denver. "It so happened that the company was in the process of upgrading its legacy system to an ERP package simultaneous with our efforts, and those were the two packages that were under consideration," he says.

"We looked at these four pretty rigorously, but ultimately decided that Prescient best met our needs," says Kahler. Key criteria included user friendliness, the ability to communicate easily with field sales people and the ability to specify what the basic stock level for products should be based on forecast accuracy. "Of course we also wanted high value for our dollar," adds Kahler. "Prescient had a better value proposition than JDE or SAP and met our functional requirements better than Mercia."

Implementation of the Prescient demand-planning package began in the summer of 2001. Kahler led the implementation teams, which also included IT and finance representation. "With commercial packages, I think it is a little easier and makes more sense for the user group to drive the project, as opposed to IT," he says. "Along with the implementation consultant from Prescient, we developed the strategy and the task list and monitored progress."

Alec Elmore, Prescient's vice president of product strategy, says that the company was sensitive to the fact that Kahler's organization already had an S&OP process that basically was working well. "Anytime you walk into a situation where there is an engrained process and you are introducing new tools, you want to make sure you are not changing the process to fit the technology," he says. "We wanted to take the best of what they had been doing for the last seven years and the best our system had to offer and marry those up to create a better solution."

The system went live in May 2002, about four months later than originally planned. During implementation the division got into "a bit of a resource crunch" and Kahler also had to deal with some unexpected turnover in the demand-planning group. "We never went into a period of hiatus but we did slow things down a bit," says Kahler. "We never worried about adhering to an artificial deadline, though. We agreed that we would go as fast as we could while staying in control and using our resources effectively. We said, 'If that means we come in three or four months after the target date, then so be it.' We didn't panic and we didn't jam it in to meet a date." As a result, he says, when the project did go live, things went very smoothly.

Users in the first phase included only the small demand-planning group. Employing the tools in Prescient's Demand Planning, they drew on historical sales figures and received inputs, via Excel spreadsheets, from product managers and field sales people. They then massaged this data to come up with final forecasts. "For the first time we were able to generate statistical forecasts by SKU and also at the brand level and the family level, which is a collection of like brands," says Kahler. "So we were able to manage the S&OP process at each of those levels."

SKU-specific Forecasts
Phase two, just completed last December, used Prescient's Enterprise Forecasting to link the system with field sales staff, so that the division's approximately 25 sales people now input directly into Prescient. "The nice thing is that the data comes in electronically and automatically is integrated into Demand Planning. We didn't have that with the spreadsheets," says Kahler. "But the real beauty is that our field sales people don't have to go to another system to get the sales history as they did before. It's all right there and it can be customized so that they only see their customers, their items, and their sales numbers." Moreover, he says, Prescient enables them to forecast at any level in the data hierarchy. "The SKU-specific forecasts that we are asking them for would have been totally impractical before." he says.

The package also enables the division to better focus on customer-specific projections. "The field sales people have the very best view of the marketplace, especially by customer," says Kahler. "So we now are asking them to give us their forecast for the next four months by major customer. Before they just gave us an aggregate forecast for their total business, but now we are saying, 'Let's look at those three or four major customers.' We give them the tool that lets then look at each one independently and that enables them to easily get the information back to us."

One of the advantages of customer-based forecasts as opposed to aggregate forecasts, says Kahler, is that it brings home to the field sales people the importance of the forecast exercise. "We take every opportunity to reinforce with our field sales people that the forecast is what determines inventory levels and, ultimately, how well we serve the customer. We explain what happens to the forecast after it is generated - that it goes to purchasing and then to the supplier and comes back as inventory that is either there to support customer orders or not, depending on how good we are. When we show them, with empirical evidence, that our ability to support Wal-Mart, Target, Food Lion and all of our major customers depends on forecast accuracy, then we have their attention. That puts money in their pockets."

The Prescient-enabled forecasts are now a key part of discussions at the company's Sales and Operations Planning meetings. S&OP is a disciplined process, Kahler explains. A series of meetings are held early in the month to the middle of the month, with each monthly cycle culminating in a meeting with the president of the Household and Body Care division. The end goal is an agreement between participating departments on the best course of action to achieve the optimal balance between supply and demand. At each stage of this process, it's important to have the most up-to-date information.

Immediate Changes
The demand plan is run monthly, but the demand planners are looking at information every day. "If they see something that is different than the assumptions we used a week or two weeks ago, they go in and change the forecast immediately," says Kahler. This could be the addition of a new customer, the loss of a customer, a competitor's price change or a new promotion. "For example, a customer might have agreed to run what we call a special pack - maybe a display case featuring two or three SKUs. As soon as we know the composition of that display pack and the volume, we put that into the forecast so that our purchasing people can see those requirements. Then, at the next S&OP cycle, we use the new numbers as an assumption."

Every Friday night the Prescient system uploads the forecast by SKU, looking out seven months. This is automatically fed to the materials resource planning system. "We run MRP over the weekend and then every Monday morning purchasing people see the latest view and take appropriate action," Kahler says. "They issue purchase orders, or move the timing of purchase orders back or forward, or they change the amount and volume on purchase orders to complement the requirements they see in MRP that are consistent with the level of coverage they have decided on for that particular SKU." The household division currently uses an MRP system from FSA, but will soon convert to the MRP module within J.D. Edwards, which the company selected as its ERP provider.

Benefits from the new forecasting program are evident in all of the household division's key performance metrics, Kahler says. "Improvement in forecast accuracy is the most direct indicator that things are working well," he says. Kahler explains that the division uses a very stringent 'absolute variance' method to assess the accuracy of forecasts. This means that the company calculates the difference between the forecast and actual sales by SKU and then drops the sign from the variance, so that plusses and minuses don't get a chance to offset. In other words, if it oversells by 10 percent in one area and undersells by 10 percent in another, both are counted as variances. "There have been many months when we have forecast a certain aggregate sales volume - say $15m for the month - and have come in right at that figure, but with a forecast accuracy of only 70 percent, so it's a pretty rigorous measure," says Kahler.

In fiscal year 1998, he says, the household division's forecast accuracy was about 51 percent. Last year this figure was up to 73 percent.

Perfect Orders
Another key measure for Kahler is Perfect Order Fill Rate. "We look at every single order shipped during the month and calculate how many of those orders were shipped complete, in terms of the line items and quantity, and whether they arrived on the customer's request date. If any of those things are violated, then we don't consider it a perfect order."

Before implementing the current demand-planning program, the company scored in the 60s on this measure, Kahler says. "Now our Perfect Order Fill Rate is in the 93 percent area," he says. "Of course, there were factors other than forecast accuracy that contributed to this improvement, but forecast accuracy certainly has contributed greatly to the improvement."

The third measure, inventory turnover, is a little more difficult, Kahler says, because of external factors that impact the calculation but have nothing to do with forecast accuracy. One is the increased safety stock that the division must keep on hand to balance the longer lead times involved in importing an increased volume of goods from Europe. Second, the finance department has changed the way it accounts for inventory. "For example," says Kahler, "the products we get from Europe go on our books as inventory the minute they are shipped, not when we receive them. At the same time, whereas we used to release our inventory the minute we made a customer shipment, now we can't recognize the reduction in inventory until the customer receives the stock. That has added two to three days of average inventory to our turnover calculations, so inventory turnover actually has gone down from four times a year to three. But if you strip away all those changes, we are really doing a pretty good job in that area."

Moreover, says Kahler, the new system is designed to enable continuous improvement. "That is why in every S&OP meeting we have, the KPI review is the very first thing on the agenda. It is very high profile, including during the final S&OP meeting each month with the president. We always talk about these three measures, what the trend is and what steps we can take to improve the numbers. It's how we figure out how well we are doing and, more importantly, how we can get better."