Forecasting demand has always been a thankless, if necessary, task. Organizations need some way to divine customer behavior - how else could they determine the right amount of raw materials to purchase, factories to build, warehouses to lease and carriers to transport the goods? Yet just about any planner will tell you that the process is accompanied by a certain amount of futility. The forecast, they say, is always wrong. It's just a question of how far off the mark you are. Now, however, companies are finding new ways to enhance the accuracy of their projections - and make up for the shortfall through a more responsive supply chain. Their newfound agility is aided by the practice of demand sensing. In this interview, conducted at the 2013 conference of the Institute of Business Forecasting & Planning in Scottsdale, Ariz., Rick Davis, vice president of business planning with Kellogg USA, talks about how demand sensing can be incorporated into traditional forecasting methods. He also discusses the value of product and customer segmentation.

Q: How is the practice of demand sensing changing the role of the forecaster today?

A: Davis: The forecasting role is evolving. There's more and more data, and technologies that allow us to look at demand signals differently. In the past, you had to make an assumption around where you thought the business was heading. Today, with daily data from critical retailers on specific products, you have the ability to sense what's going on in the marketplace. That changes the role of the forecaster, because he's sitting on very powerful information. It's less about making a judgment call about where you think demand is going to go. You're grounding it in reality, understanding where demand is, then working that trend forward to make certain that you're starting and ending from a better place.

Q: We hear so much these days that the forecast is always wrong. It almost leads to the attitude of why bother? Why not just focus on agility? It sounds like demand sensing is a good way to approach the problem.

A: Davis: It is. It gives you a better starting point, and I think that's what is critical. When you talk about having a flexible supply chain, or the ability to respond to demand shocks, demand sensing means that you need to think about that differently. If you get the signal and you can't do anything about it, you've gained very little. So it's about the organization being willing to change - to take that information and then work differently. You have the ability to do a much better job on deploying product at the right place and the right time, to get it in the customer's hands. If you haven't built infrastructure to respond to that appropriately, then you're going to miss the opportunity.

Q: Should we be thinking about demand sensing as something that's tactical in nature, or strategic?

A: Davis: I think it's both strategic and tactical. The organization being willing to work in that manner is absolutely a strategic decision. Tactically, how you respond is a different conversation. In the very near term, it's a much better way of understanding what's going on, and whether my inventory strategy can support me with what's taking place today. Can I have the infrastructure created to respond and actually manufacture to that pull trigger? We're integrating more appropriate trends into our statistical models, and learning as we go in a much quicker time line. The latency is far less than what we've been accustomed to in the past.

Q: We've heard a lot recently about a "one-number" planning process, almost to the point where it's become gospel to some companies. Is that truly obtainable?

A: Davis: I believe it's a bit of a myth. You'll hear a lot of people talk about the one-number process, and what that really means. Many have said it's much about alignment, and I agree with that. It means the organization needs to be marching to a common plan. I also think it gives you a basis from which to manage exceptions.

If we start from the same place, and see what changes over time, it's a little easier to manage the exceptions. The number moves continuously. It's more of a snapshot in time, an opportunity to drive alignment. The reality is if we that we're not going to respond to change, we're going to make a lot of customers very unhappy.

Q: It must be a delicate balance. On one hand you have to allow for the fact that there are going to be changes in the plan. At what point is there acceptable wiggle room? And when is it simply throwing the plan out, making it useless?

A: Davis: That comes back to making sure that the plan is grounded in fact, data, logic and reality. Nobody has a crystal ball. I don't believe we can just say that looks low or too high. Any change that we make to the forecast - any variation from the starting point - should be because we have new news or better information.

Q: One of the complexities that has been introduced into the discussion in recent years has been the idea of product and customer segmentation. How do those two work together in order to create a better forecasting and planning environment?

A: Davis: This isn't a new theory; many have tried to do it. We've had it built into some of our processes along the way. We all know which products and customers drive the volume. It's a way of working differently, to make certain that we are aligned and we have the same priorities. It's also important to understand that each product has characteristics and behaves differently. So whether it's an innovation item or product at the end of its lifecycle, or one that's fairly consistent with low variation, it's about understanding the different characteristics of those products, so we can apply the right demand signal to give you the best predictive path.

We find that for some products in that stream, we don't need to do a lot at all - the statistical model can give us a pretty good path. There are others where we have to intervene quite frequently, and we need demand signals from the retailer, trade inventories or some other venue, to make sure that we're making the best choice around what signal to select to lead our forecast. It's really about applying the right information to the right product.

With customers it's the same way. It doesn't mean that one customer is less important than another. What you need to understand is how customers are variable in their promotional plans - the things they do that are different, that drive demand shocks within the system. How can you be ready for those intersections of customers and products that can drive significant variability? You want to factor them into your inventory strategy, and the way in which you set customer-service targets around certain products.

Q: I could see you slicing and dicing so thinly by product and customer that you'd end up with so many different sales and operations planning processes that it would become chaos.

A: Davis: It requires a lot of discipline. Most organizations struggle significantly with that piece. Even at Kellogg, we've been better at this in some areas than in others. It's about making the choices and then sticking to them. It's also understanding that product and customer segmentation is evolving. You have to take a position, make a start. You might use two or three specific characteristics where you want to make your first cut at it. Over time, you'll find that it evolves, and that you have to add more and more. The discipline lies both in getting started and in keeping it from growing into something that you can't manage.

We don't need more complexity - we actually need less. If you do product and customer segmentation appropriately, it should take out complexity. It should help you understand the products on which you should be focused, the ones that are highly promotive, and the ones where you might want to do customization. By understanding both the inventory and distribution strategies around that, you can provide a better service rate.

To see the video of this interview, visit SupplyChainBrain.

Resource Link:
Kellogg


Keywords: supply chain, supply chain management, inventory management, logistics management, warehouse management, supply chain planning, retail supply chain, supply chain risk management