Visit Our Sponsors
Syngenta is a world leader in the production of crop-protecting herbicides, fungicides and insecticides. It also is the world's third-largest producer of high-value commercial seeds. The company employs some 19,000 people in more than 90 countries and generated sales in 2004 of $7.3bn. It recently implemented a forecasting solution from John Galt of Chicago.
Q: How did you come to be involved in supply and demand management at Syngenta:
Herrin: My background is actually in chemistry and chemical engineering. I earned a chemical degree from the University of Jacksonville in Florida and a chemical engineering degree from the University of Florida. So I started working in the plants, doing project and process engineering. Over time I moved into production engineering. Then, when the company decided to start a sales and operations planning (S&OP) process, I was named the S&OP manager for the fungicide/insecticide parts of our business. S&OP later expanded to all of our North American operations and I became the head of S&OP for North America. However, to effectively run S&OP, you really need a good forecasting process and we didn't have that. We had the planning part, but not the forecasting part. So, over time, we started looking for the right forecasting solution and I moved into my current position, where I am charged with developing a forecasting process.
Q: I know that you are responsible for the NAFTA region. How does this region fit into your global supply chain?
Herrin: We source globally and have plants all over the world that produce our active ingredients. The active ingredients are shipped into the regions and the regions formulate and package them into a variety of offerings. Each region customizes the end package for its local markets. What distinguishes the portfolio are the brands and the ways we package them. Our distribution channel is pretty standard-we sell to distributors who sell to dealers who sell directly to growers. There may be a few exceptions to this, but that is mainly the way it works.
In the NAFTA region, we have two active-ingredient plants in the U.S. and a variety of formulation and packaging units around the U.S. and Canada and Mexico. The majority of our active ingredients are sourced globally from Europe, India and China.
We are involved in all kinds of shipping. Imported active ingredients are shipped by boat or plane. Finished product is moved pretty much by truck and we do have a lot of rail movements.
Our products are extremely seasonal. In the agriculture business, herbicides, insecticides and fungicides are pretty much applied in very narrow windows during the growing season. For most products we peak over a three-month period and then the rest of the year is intermittent demand.
Q: What were your primary challenges in terms of forecasting?
Herrin: When we started, we really didn't have a forecasting process. We had a planning process. We would take our sales targets and blow them down. But when we had our sales and operations planning meetings, we never got a picture of market reality-of what was coming up. So the first thing we did was to ask people at lower levels to start forecasting. They would send in an Excel spreadsheet once a month, but there were a lot of problems with that. It was not a live process and it didn't give us visibility to all the other factors that go into forecasting-people would just enter a number for a given product and they pretty much gave us back our target number. So we still weren't getting a true picture of what the market was doing. Also, we had a pretty cumbersome, once-a-month process of e-mailing Excel spreadsheets around. That was confusing because we never knew which version of the spreadsheet was the latest one. So our issues were really visibility of supply and visibility of historical sales. We just didn't have the things you need to develop a true, bottom-up forecast. People had to go out to all kinds of different sources to get this information.
Q: How did you go about selecting a forecasting application?
Herrin: One of the key things we liked about the John Galt ForecastX solution was its Collaborator module. This is a web-based collaborator that everyone can use to input what they think their future sales are going to be. They can very easily see it in units, in gross revenue, in net revenue. They can also see what everybody else's opinion is for the same region or product. So somebody at the district level can put in his forecast and he can see on a separate line what his supervisor expects him to sell. Before, we didn't see that gap. We had the one number, which basically was the targeted number. Now we can separate the targeted number from the forecasted number and measure the gap.
So, I see my budget month by month for this product; on the next line I see what my supervisor wants to agree to for this product by month; and then I can enter what I actually think I am going to sell, based on my current assumptions. We measure the gap monthly and we close the gap by addressing the assumptions.
Q: Can you give me an example?
Herrin: The guy at the lowest level says, 'OK, with my current competition, my current pricing and current weather conditions, I think I can sell 100 units of product X over the next three months. His supervisor or sales manager may look at that and say, 'that is short of what I want to commit to as revenue for the next three months.' So he can put in a different number, bring it up or down. If he brings it up he is doing it on a separate line. We don't overwrite that original forecast. Then we show the gap. Maybe the district guy forecasted sales of $100m, while his supervisor said, 'no, I have to hit $110m.' That leaves a $10m gap. Now they can argue about how to close the gap. The regional guy may say, 'If you want an additional $10m, which assumptions are you going to change? Are you going to give me a pricing break or give me more advertising? How are we going to change the assumptions to close that gap?
If they manage the assumptions correctly, that gap will close. If they don't manage the assumptions correctly and if the guy is still giving his unbiased, true feeling of what the bottom-up number is, then we know that the assumptions we are playing with are not true market drivers. Hopefully, in the long run, it will give us an opportunity to really focus in on what the true demand drivers are for our business. And if we run a program or discount, we will be able to go back and see if that program we targeted actually made a difference in demand.
Our company still has a cultural bias toward the financial forecast and that's hard to break. But I can see the S&OP process starting to change that. At our S&OP meetings we are starting to make decisions about the marketing levers we have and how are we can change those to close the gap between those two numbers.
Q: Does the S&OP meeting take place online?
Herrin: We actually have formal pre-S&OP and S&OP meetings. But this could be done online because all the data is in the Collaborator, which is part of the John Galt Atlas suite. I call it a spreadsheet on steroids because it gives you all of the functionality of Excel and allows you to do filters and aggregation. The software resides on our intranet and everybody is given an account. The administrator tells each person what products they can forecast against in which regions. Then, in the tool, we can actually aggregate across all of our sales people up to a business-unit level or national level.
That then drives everything else in the supply chain and purchasing. The Collaborator is where we capture the sales forecast, the sales plan and marketing's request for product and this data actually feeds our other systems, including SAP APO.
Q: Is your primary goal with this solution to improve forecast accuracy?
Herrin: Our goal is to drive and improve forecast accuracy and then to drive assumption and event management, which is really part of improving the accuracy. Right now, we are in the process of implementing statistical forecasting using John Galt. Statistical forecasting pretty much says, 'If you knew that in the future you would do about what you did in the past, this is where you would fall out-your range. So we are giving people a statistical range. And then we have a manual process when people go in and manually input their numbers. Anytime they are outside of that statistical range, we are going to drive an assumption or an event. We go out and tell them, 'OK. the range says that you should be here. If you are outside that range, obviously you are going to do something different or something different is happening in the marketplace. We need for you to tell us what that something is.'
In the end, you get a much better understanding of the market you are working in. It also gives us a base line against which we can measure our promotional programs. If we are going to run a program, a sales or marketing person might say that it will give us a 20 percent lift. We can go back and say, 'Well, no. Based on our statistical base line, it gave us a 5 percent lift.' We can actually measure how effective our marketing levers are being managed.
Q: What led you to select John Galt over other solutions?
Herrin: Most of the other vendors we talked with wanted to come out with a full-blown statistical system-almost an ERP implementation. But since we were moving from spreadsheets, we wanted to make the transition as painless as possible for the people who were actually doing the forecasting, most of whom are not technical experts. Most of them are technophobes. They were comfortable with Excel spreadsheets and, with the Collaborator, Galt gave us a simple replacement for that. The difference is statistical forecasting vs. consensus forecasting. We wanted to start with consensus forecasting because we weren't ready for statistical. We just didn't have the skill set in-house, or the knowledge or the faith that our business could be statistically forecasted. So Galt allowed us to immediately replace the Excel system that we were manually e-mailing around with something that looked just like the Excel screens.
Then as we got more informed and better at the consensus, we decided to go ahead with statistical. We first used the Collaborator just to collect sales and then we started doing supply balancing, taking marketing's product requests and balancing those requests against the sales plan. We started using the tool to show the guys how much extra inventory they were asking for and we actually built performance measures for supply-and-demand balance. This was stuff that we didn't have with Excel. So the Collaborator gave us a live system and it took all of the workload out of moving these Excel spreadsheets. We still had the same amount of workload in the forecasting end-the users still had to go in and put their numbers in monthly. But at least now we could pull back a history. We could show them, on a line directly under their entries, their prior years' history and the budget. So we are able to give them a little guidance.
The Galt guys came in and under two months had this system up and running fully. And they did it for what we thought was a very good price, way under what everybody else we talked to wanted. As I said, the other vendors were pushing a complete statistical forecasting system, but we just didn't have the staff or the expertise in-house to support that. The Galt guys push a "walk, run, fly" methodology. We needed to walk at first. Actually, we needed to crawl at first.
Q: Are you walking now?
Herrin: I would say we are almost running. Once we get the statistical up and going, I think we will be ready to fly. At that point, we will be running our whole planning system using statistical guidelines. On the forecasting end, the only time somebody is really going to have to go in and play with their forecast is if they know something different is happening. It will be almost a complete exception management system. We are actually going to quadrant out our products. Those that are highly forecastable, we will let the system handle. Stuff that isn't highly forecastable, we will say, 'OK, here are the handful of products that you guys have to pay attention to and that you really have to manage month to month. If it is highly forecastable, we will do it statistically and manage by exception. If the actual comes in outside of our statistical confidence interval or if you see something happening that is going to significantly change the market, then you have to go in and do something.' And for products that are important to us and are very unforecastable, we need to start asking whether we should convert these to make-to-order products or whether we should even be in that business.
Q: Do you have any quantifiable results or "lessons learned" that you can share?
Herrin: A lot of this is still in progress. We certainly have made inventory reductions over the last two years. But whenever you show a success like reducing inventory, how do you attribute that back to the improved statistical forecasting? Most groups will jump up and say, 'we reduced inventory because I planned better or purchased better.' What I tell everyone is, 'hey, these are the numbers that all our plans are based on, so if we didn't improve these numbers, you could not have gotten it right in the supply chain.'
As for lessons learned, our biggest lesson was that there are three focus areas for implementing something like this. One is the tool, one is people and the other is process. You can't just go out with a tool. You first have to have a process that is documented, lined out and agreed to. If you have the process developed, then it is easy to automate using the tool, but you can't depend on the tool to drive a process. That is not going to happen and you don't want that to happen because then you are changing your process to match what the tool can deliver. The first thing is to really focus on process.
The biggest challenge we have, though, is on the people side-changing behaviors, identifying capabilities and aligning those capabilities to deliver what we need to deliver. That is the tough part. If I could start again, that is the thing I would address first and I'd really put a lot of effort into it. You have to change people's behaviors. We are still fighting the bias people put into forecasting their financial numbers. Sales guy will always under-forecast and marketing guys will always over-forecast. Without addressing that upfront, really working with people and convincing them of the importance of coming in with a really accurate and unbiased number, the rest doesn't make sense.
Enjoy curated articles directly to your inbox.