Building a Forecasting Process from the Ground Up

Moving to reduce inventory while maintaining customer service levels, agribusiness Syngenta needed accurate demand forecasts. But first the company had to put a forecasting process in place.

Moving to reduce inventory while maintaining customer service levels, agribusiness Syngenta needed accurate demand forecasts. But first the company had to put a forecasting process in place.

Forecasting demand, as much an art as a science, is never easy in the best of circumstances. But Greensboro, N.C.-based Syngenta Crop Protection faced at least one major hurdle in forecasting North American demand for its fungicides, herbicides and insecticides, according to Richard Herrin, manager for NAFTA planning and forecasting at the company: "We didn't have a forecasting process."

Now the company has undertaken an initiative to implement just such a process and to provide its employees with the necessary forecasting tools as part of Syngenta's drive to maintain its customer service levels while driving down its working-capital requirements.

Mostly Seasonal, With Variable Demand

Syngenta Crop Protection is a $2.3 billion business in North America, based on 2002 sales, and a $6.2 billion business globally. With 19,000 employees in 90 countries, including 4,300 in the North American Free Trade Agreement (NAFTA) region, the company is headquartered in Basel, Switzerland. Syngenta sells its fungicides, herbicides and insecticides to distributors and dealers, who then sell to growers.

The agribusiness market, and Syngenta's segment of that market in particular, offers a challenging environment for demand forecasters. Syngenta produces the active ingredients (AIs) with a two-year lead-time. The company distributes these AIs to various regional divisions, which formulate and package local variations of its products (stock keeping units — SKUs) in accordance with the local requirements and regulations. The challenge for the local units has been to manufacture enough of each particular product to meet demand, even though, absent a formal forecasting process, the production side of the business did not have a clear picture of actual demand.

In the past, Syngenta employed a "top down" business planning process for generating sales goals and production targets. The company's finance group would generate quarterly sales estimates and set targets for the sales staff. On the manufacturing side, the company's production planners would blow out those targets to determine how much of each SKU to produce. The catch, of course, was that without accurate demand forecasts, the company found itself making to stock and loading up on inventory to ensure that they could meet fluctuations in customer demand and minimize orders lost due to out-of-stock product. Inevitably, the mix of goods that Syngenta produced in line with the sales goals would not be the mix of goods that wound up selling, and at the end of the January-May production cycle, the company typically found itself with excess inventory. "We had huge inventories out there," Herrin says, adding that the company's inventory turns were about one-and-a-half per year. And it's no wonder, given the highly seasonal nature of the company's sales and production cycle, the agricultural market's perennially high demand variability and customer delivery expectations of less than one week (and frequent cases in which two-day delivery had to be guaranteed).

On top of these challenges, Syngenta, like other companies in this extremely competitive segment, faced the imperative to minimize its working capital needs while maintaining its high customer service levels. To get better control of its inventories, Syngenta elected to undertake a major sales and operations planning (S&OP) initiative. Trouble was, S&OP uses inventory control models to affect working capital requirements, and the basic inputs into the inventory control models are accurate demand forecasts, but, as noted above, Syngenta did not have in place a forecasting process to gather those inputs.

Spreadsheets Offer Clouded Visibility

The company made an initial move to get at forecast numbers using the favorite tool of many a planner: Excel spreadsheets. Once a month, Syngenta's central office would create a spreadsheet for each business unit, with volume and price forecasts, and e-mail the document to a contact within the unit who had been tapped to do the forecasting. The business units would revise the figures and e-mail them back to the central office, where another staffer combined all the numbers for manual uploading into the company's enterprise resource planning (ERP) system.

This approach had several shortcomings: First, the company was updating its forecasts only once a month, but changes in the market might require daily revisions to production plans. In addition, the spreadsheet forecasts reflected sales estimates made by business unit heads and excluded insights that could be provided by field reps and their local district managers, that is, those employees who were closest to the customers and who could offer valuable perspective on market trends. And finally, this manual process did not allow the company to identify and manage gaps between the forecast and sales plan, and forecasters did not have ready access to such information as sales to date, current inventories or marketing plans.

Around the beginning of 2002, Syngenta began looking for a more effective tool for forecasting that would overcome these shortcomings. The company's requirements included, among others, that the tool support Syngenta's S&OP process cycle at the business unit, country and NAFTA levels; allow gap, exception and assumption management; allow for multiple languages and currencies; and meet the needs of the various functions that would be using the forecast, including supply, purchasing, sales, marketing, finance, development and strategic planning.

Identifying the Right Tool for the Job

With that broad mandate, Herrin and his team surveyed the market for collaborative forecasting engines and eventually settled on a solution from Chicago-based John Galt Solutions, a privately held company that the Syngenta group had met at a conference of the Institute of Business Forecasting. Herrin says that Galt's ForecastX solution was particularly appealing for its adaptability. "Everybody else we talked to was going to come in and tell us, 'You have to forecast in this way,' because that was the way that their software worked," Herrin says. On the other hand, "Galt was very flexible, and we could do a lot of front-end customization with their front-end engine."

That was important because, as Syngenta began implementing Galt's solution in mid-2002, Herrin and his team wanted to make the transition to the new tool as painless as possible for staff in the field. "We had to come out with something that would say, 'Hey, just tell me what you think you're going to sell,'" Herrin says. "So we actually designed it to look exactly like the spreadsheets that the guys had been e-mailing back and forth."

Galt also scored by coming in with a project of a scale and price that Syngenta found acceptable. "Cost was a big issue, and Galt came in lower than anyone else," Herrin says flatly. "Everyone else we talked to wanted to come in with a big package, a full-blown forecasting tool. And we told them, 'We just need a collaborator, something where everyone can put in their opinion and we can aggregate it up.'"

Currently Syngenta has about 250 users working with the forecasting tool, although the company uses a solution from Cognos as the reporting tool for Collaborator, and essentially the entire company can access the reports that the forecasting tool generates. Collaborator also feeds Syngenta's ERP system with sales plan and supply requirements. "So if you think about it, the whole company is working off numbers that are fed directly from Collaborator," Herrin says.

A New Process

Syngenta's formal S&OP process, as it stands now, works like this: At the beginning of the month, the company's crop managers (CMs) — one step up the organizational chain from the local district managers — enter their sales forecasts for each SKU. Those forecasts get rolled up to the business unit level, and by the end of the first week of the month, the business units develop, and commit to, their own forecasts, which may or may not conform to the CMs' forecasts (both are shown in the Collaborator report). Once all those data are entered, in the second week of the month the brand managers — the next step up the organizational chain — review the sales and demand forecasts, and then they input their product supply requirements, which are fed into the ERP system and used by the company's supply chain function to plan production.

In the third week of the month, the company's S&OP meeting brings top executives together to review the current supply and demand plan against the annual budget. Based on any trends occurring with supply and demand, or in the marketplace, the S&OP meeting could produce changes in the sales plan, which are then entered into Collaborator. Finally, toward the end of the month, district managers enter their seasonalized invoiced sales forecasts into the system, in part based on feedback from sales reps out in the field. The district managers' input feeds up the chain to the crop managers, and the process begins again.

The forecast horizon varies for the different participants in the forecasting process: for district managers, it's 12 months; for crop managers and business unit managers, it's 12 to 24 months; and for brand managers, it's one to five or more years. Currently, Syngenta is forecasting its entire NAFTA region sales volume through the Collaborator solution.

Putting the Process Before the Tool

Implementing the new process, and the new tool, has not been entirely without challenges, both technical and cultural in nature, according to Herrin. On the technical side, the issue was data accuracy in the various systems from which the forecasting engine pulls information, including the financial, ERP, manufacturing resource planning (MRP) and supply chain planning systems. Herrin says that in the initial phase of the project, the implementation team spent a good deal of time working with the caretakers of these other systems to ensure that data were up-to-date and would remain so moving forward. The effort was worth it, Herrin says, because ultimately the forecasting system is providing a central point where anyone in the company can see all the information that they need to do their job. "Instead of going to 50 different reports, they can go to one screen and look at all the information," he says.

On the cultural side, the challenge has been ensuring that all interested parties within the company buy into the new process and adopt the new tool. To ensure adoption, Herrin says that his team has come to take a process-based approach rather than focusing solely on the new solution. "A tool is just a way to automate a process," he says. "If you don't have the process implemented or lined-up to begin with, the tool is never going to be successful." To some extent, that approach has extended the roll-out period for the new tool, because the company has had to customize both the process and the tool to fit the unique organizational structure of this or that business unit. Over time Herrin's team also has had to take into account changes in business units or in the sales force's structure, requiring that they readjust the process or forecasting tool to keep up with the changes.

Despite the challenges, Herrin says that the forecasting initiative appears to be paying off already in that, combined with other project initiatives, the company has been able to achieve inventory reduction targets and has focused attention on tracking and reducing top-line and product mix forecast error. Syngenta also has been targeting an increase in inventory turns, but Herrin was unable to say at the time of the interview for this article whether that result would materialize to the extent hoped. The other key metric the company is tracking to judge the success of the forecasting project is reductions in lost sales; but, again, it remains too early to say to what extent the initiative is paying off in this area.

Unexpected Collaboration

One unanticipated benefit of moving to the new process is that the company's commercial side, per force, is working more closely with the supply chain side. "Commercial can now see what supply chain is doing, because we have a supply-demand balance calculation in the Collaborator," Herrin explains. "Making that balancing process visible is prompting those two groups to work a lot closer together." For example, the new process is prompting supply chain and commercial to identify and share risk to an extent not seen in the past. Previously, for instance, if commercial had an upside of 20 percent they would put in a request for the 20 percent, and the company would make it. "Now supply chain can come back and say, well, if this is a risk of 20 percent, instead of me making the inventory, maybe I could hold flexibility," Herrin says. "Maybe I don't have to commit to the inventory. So we're better balanced between supply and demand plans."

Looking down the road, Herrin says that the company is exploring using the new process and Galt's Collaborator solution to support the entire S&OP process as a sort of executive information system, to give the participants in the S&OP meeting all the data they need, aggregated to the appropriate level, when they need it. In addition, Syngenta is looking at implementing Galt's ForecastX SDK solution for statistical forecasting to provide baseline performance metrics. Explains Herrin: "Once we have a baseline, we can use that to drive assumption management. Currently, if someone says they're going to launch a program and get 10 percent lift, we don't have a baseline to measure that against. We don't know if he got a 10 percent lift or not. If we had a statistical baseline to measure that against, then we could tell him exactly what his lift was."

Elsewhere, Syngenta is looking to use Collaborator to drive distribution and logistics planning. And finally, the company is looking at the possibility of extending the collaboration aspects of the new forecasting process out to both its customers and its suppliers. "I want to eventually take the Collaborator to our customers and say, 'Give us a forecast,'" Herrin says. "And also to share the variability with our vendors. I want to say, 'Look, here's our variability around the market, what can you do to help us?'"