There are so many parts - and people - in the supply chain, that top performance can be achieved only when all silos are broken down and a truly holistic view is taken, from upstream to downstream. Ideally, that means everyone should have some idea of the role and function of every other person in the enterprise, or at least of their department. But does that mean participation in other units' work should be encouraged? Theoretically, demand planning accuracy would be improved if internal stakeholders, such as sales, marketing, supply chain, even C-suite executives, made their contributions to forecasting. But is that the case?

SupplyChainBrain editors interviewed a number of demand planning experts on the pro's and cons of that, and on many other topics touching on the discipline. Extracts from those conversations, held in October 2015, at the Institute of Business Planning & Forecasting's conference in Orlando, are compiled below.

Q: Let’s talk about data availability and analytics and how they have transformed the demand planning landscape.

Manikantan Aryapadi, manager, A.T. Kearney: I would characterize the growth of analytics and data in three dimensions: volume, velocity and volatility. There has been tremendous data creation in recent years, with 90 percent of that data created in just the last five years.

Volatility is the frequency that we get the data – either clean or dirty – along with all of its components and sub-components. Variability is a big component of that.

Historically, companies didn’t have access to this data because technology had not evolved to help them generate actionable insights from the data. Now everything has changed. Over the last couple of years, a lot of tools on the analytics side give companies the ability to analyze what I call two sets of data: structured and unstructured. Structured is where you have such things as point-of-sale information or customer-related information, that’s codified and classified, and the volume is huge. Unstructured includes things like weather patterns and traffic.

The ability of analytics to sort of marry both structured and unstructured and provide actionable insight to companies is transforming demand planning.

Q: Can you give us an example?

Aryapadi: How we see that changing the landscape is this: if it's going to rain, you may want to know a day or two in advance, can I have Walmart ship an entire set of umbrellas to stores where they expect consumers to come in. The same goes for beer or cereal.

You gain an understanding of what actually drives category planning and how to better place your items. From a demand planning perspective, it’s all about placing the right product at the right place at the right time in the right quality. That's the holy grail they are chasing. Now we have the ability to analyze this information, present it, and deliver actionable decisions from the allocation standpoint and from the supply chain standpoint. We have the ability to actually be able to have inventory in the store at the right time.

Q: Are we talking about an ideal state versus what’s actually going on in the real world? How effectively are people actually using the data available in their forecasting?

Aryapadi: What we’ve seen is, clients are using it but not as effectively as they should. They are using it in bits and pieces. The power of demand planning, integrated with your supply chain, can only be harnessed if you can look at it holistically and entirely. When you're doing bits and pieces from one segment of the supply chain, point of sale, for example, but you aren't marrying it with larger structured and unstructured analysis, that won’t give you actionable insights. Some companies are beginning to do that, but we're only at the tip of realizing the potential. We will get there in the next four or five years.

Q: How important is SKU-level forecasting?

Aryapadi: I think it’s the purest form of demand, which means it is the closest to the consumer. For example, when you go into a retail store to buy a pair of jeans, you're not just buying jeans. You're looking for a particular fit and a particular size and a particular color. That’s what SKU-level information is. You want to make sure that companies tailor their inventory to your consumer preferences.

Again, historically, this data has always been available because POS systems have captured it for many years. We've seen companies that have 10 to 15 years' worth of actionable data, but it just sits there for a lot of reasons. It's difficult to analyze, for one thing. The talent pool did not exist to analyze it. Processes were antiquated and old fashioned. From a technology standpoint, a lot of tools used in the ‘90s, even in the early 2000s, did not have the ability to analyze this information at the SKU level. This has seen a major change in the last five years. Data scientists have been created. You have a new group of people, extremely talented, strong in analytics, able to analyze almost terabytes of information, using the right tools, and feeding the data to allocation systems so they can digest it and make changes accordingly. That's why SKU-level forecasting is more important now and can give you competitive advantage. 

Q: You’ve written that a company can no longer hide behind MAPE, or mean absolute percentage error. What did you mean by that?

Charles Chase, advisory industry consultant, SAS:  As demand planners we tend to focus on MAPE to measure our effectiveness when in fact consumers of our forecasts look at things more business-relevant like revenue and profitability. We tend to focus on what we're going to forecast rather than on the things that influence the forecast.

Q: How do we change that?

Chase: In order to understand that, you have to invest in more than just process and technology. You have to invest in people skills as well as predictive analytics. Most companies are just using descriptive analytics to do reporting. To really understand the influences on demand – like price, advertising, sales promotion – you have to have different skill sets and technology. You have to be able to do predictive analytics, to measure impact and be more business relevant. The reason why a lot of our forecasts are not used by people like marketing, sales and senior-level management is that they are not business relevant to them. They don’t care about MAPE. They care about how much revenue will be generated based on things that will influence and affect the forecast.

In fact, most executives are willing to give up some points of accuracy or precision in a forecast for more precision in measuring what influences the forecasts.

Q: So, what are key investment areas, in your view?

Chase: People skills and behavioral changes. Also, process has to be not vertically but horizontally integrated. Many of our processes are vertically integrated and not connected. You can’t synchronize demand and supply as a result. You need to have a horizontal process with horizontal metrics. You have to share certain metrics. Sales and marketing don't measure anything by MAPE, but by performance, revenue, profitability, how much they can get for every dollar they spend. We need to find common metrics. We need to invest not just in descriptive analytics but in predictive analytics. We also need to upgrade technology to be more scalable so it can handle big data and utilize the power of analytics. So, invest in four areas: people, process, analytics and technology.

Q: A common belief is that input from the sales department benefits forecasting process, but you challenge that. Why?

Michael Gilliland, product marketing manager, SAS: Early in my career, working for a large consumer products company, I had this experience: we had about 500 finished items shipped and sold to 10 distribution centers, so we had about 5000 weekly forecasts to generate. We believed it was a good idea to get sales people involved, to provide their input to fine-tune these forecasts and make them more accurate. We went to great effort to persuade the vice president of sales to commit his people to participate in the process. We started tracking the numbers and after a couple of months of weekly  forecasting performance reports, we realized we weren’t getting any better at fulfilling demand at the DC item level, which is what we cared about.

This was a rude awakening for those of us who had pleaded for this input. We realized there were three things that needed to happen before sales input can be relevant and useful to forecasting. One, sales can actually forecast their customer demand at the item level. Two, even if they can forecast accurately, will they give an honest answer in the process, and three, even if you get better forecasts at the customer-item level, does that make a forecast that matters at the DC item level much more accurate? Those were early learnings.

Q: Assuming you want sales team input, how do you go about getting it?

Gilliland: One way is to just throw out a blank spreadsheet to sales people and say fill it in: what are your customers going to be ordering over the next weeks, months or whatever time period you're dealing with. The other is strongly preferred from the corporate forecast standpoint. You send out your initial forecast to the sales people and say this is what we think your numbers are for these customers for these items. Pencil in a change if you feel the number is different. Give them something to start with. Remember, sales people are not hired because they're forecasters. It takes a little bit of effort for them to create numbers out of the blue.

Q: In talking about the future of supply chain, what will be demanded, if you will, of demand planning?

Eric Wilson, director, demand planning and S&OP, Tempur-Sealy International: We’re talking about being demand-driven, becoming more demand-oriented toward the supply chain. Big data is getting bigger and everybody wants more, faster. But all those things require different skill sets. We have a perfect storm brewing, with massive amounts of baby boomers going out of the work force that are not going to be replaced easily. Those coming in are without skill sets for what's required in the supply chain. Companies and academics are not keeping up with this: demand will outstrip the current supply of people.

Q: How do we combat this talent drain, this so-called perfect storm?

Wilson: We have to be more creative. For instance, we may find good talent in other functional areas in our organization. Marketing can become excellent demand planners – people in sales or other areas you wouldn't typically look at as analytical. We can create those skills in them. But retention will be key. If you have 10-to-1 deficit, as some studies predict, the traditional stair step from analyst to senior analyst to manager – that won’t work in the future.

Q: How can forecast value-add analysis improve your forecast accuracy?

Erin L. Marchant, senior supply chain analyst, global demand management, Moen Inc.:  There are many different players in a forecasting process or cycle, and FVA is the tool that will allow you to assess how well each of those players contribute to the success of your forecast. There is a lot of pressure to get that forecast right, and there are a whole lot of people with input into what that final forecast should be. FVA shows who's having the biggest impact within your supply chain, and allows you to go after those parties who don't positively impact things.

If you have, for example, a particular contribution from executives or from the sales force that is actually devaluing your forecast, you can start to have one of those critical conversations about how to improve that part of your process. It’s also empowering to the demand planner. They’re  the ones who take the brunt of the beating when the forecast is bad. But with FVA you can accurately pinpoint who contributed to the devaluation of your forecast.

Q: Is everyone's input really necessary?

Marchant: I think everyone needs to have a seat at the table and make sure that they are heard in the process. You can get good intelligence from sales, sometimes even executives have good things to add to the forecast. However, I think we need to be very wary of what one's role is in that process. For example, a sales guy is always going to say he will bring in tons of sales, but that may not be the reality. So being able to segment that information and accurately measure what everyone brings can only help your process.

Actually bringing to light how they are valuing or devaluing each of those steps helps you have the right conversation. Often the conversation boils down to, well, you need better software or more information or more analytics or more people, and sometimes that’s not the case.

Q: Can a company use planning to lower its labor costs?

Billy Duty, director of the supply chain center of excellence at Momentive: Yes, there are two ways. One is being able to improve visibility between supply chain and manufacturing, and, second, integrating a part of the planning process by using block scheduling and by putting restraints on what we schedule.

For instance, a department may have five machines, but we only want to schedule four of them at a time. That reduces overtime and the labor components and we can move forward with lower costs.

Q: Momentive is in chemical manufacturing. What have been the approaches that you have used there?

Duty: Over the last few years, we have put in a finite scheduling program. We embedded things like block scheduling, looking at our capacity and saying, okay, if we’re 40 percent loaded for the month, that’s a lot of free days – let’s consolidate those days together. That way we can reduce overtime and labor on the other days. We can systematically put in the schedule that we’re only going to run four machines at a time. We put that in our finite scheduling program. Now that it's embedded into our scheduling, we’ve got visibility, and supply chain and manufacturing can work together. We’ve seen tremendous reductions in overtime and labor costs.