Every supply chain program, good or bad, launches from a plan. It's the ability to forecast and analyze product demand, consumer buying patterns, and economic trends that separates the winners from the losers. In reality, any kind of a forecast is going to involve the black arts of predicting the future, a process that inevitably will result in some errors even under the best circumstances. It's not an issue of what happens if a forecast goes wrong—it's more an issue of by how much.

The history of supply chain management includes some notoriously bad plans—plans so far off the mark that they've become legendary in the “what-were-they-thinking?” category. The bigger the company is, the more spectacular are its supply chain glitches since the ripple effects can extend well past the four walls of the company to include suppliers and customers. And as we saw in the previous chapter, the bigger the supply chain glitches, the bigger the hit a company takes to its share price and ultimately, its bottom line.

The main reason companies struggle with their forecasts is the fickleness of the marketplace. Try as hard as they might—and they've been at it for centuries—manufacturers and retailers still haven't been able to consistently figure out exactly how much of a product consumers are going to buy. Accurately forecasting product demand is probably the single most important—and most challenging—measure of a company's supply chain proficiency. Improving forecast accuracy has gotten a lot of attention, but as meteorologists have always known, you can be right most of the time, but it's the one time you're wrong that gets a lot of people upset.

When analyst firm AMR Research (which has since been acquired by Gartner) studied forecast accuracy at several dozen manufacturers, it turned out—not surprisingly—that errors are very much a fact of life within the supply chain. Forecast errors at bulk chemical producers, for instance, range from 10% to 24%, for a median error rate of 11%. That's actually pretty good, though, since consumer goods companies get it wrong from 14% to 40% of the time, or an average 26% error rate. Consider that for a minute: One time out of every four the forecast is wrong. It's even worse in the high-tech arena. The error rate ranges from an outstanding 4% to a horrific 45% rate (with a median rate of 28%). That's right—at some high-tech companies, they're getting it wrong nearly half of the time.¹

Case in point: Some years ago, Cisco Systems Inc. had a royal doozy of a glitch, one that has become the stuff of supply chain legend, so to speak. Cisco's glitch was centered squarely on the failure of its supply chain plan. As the leading manufacturer of networking routers and switches, Cisco was one of the most influential companies driving the dot-com boom of the late 1990s and early 2000s. In the spring of 2001, Cisco was riding as high as any high-tech company had ever ridden, having reported a profit for 40 quarters in a row. With a culture that literally knew nothing but growth, naturally enough Cisco's planning systems—which were considered state of the art—kept forecasting more of the same.

Unfortunately, the inevitable bursting of the dot-com bubble happened to coincide with a severe slump in the telecom industry, both of which had a direct impact on Cisco's business. The decade-long uptick had finally peaked, and demand for Cisco's products began to slow. Problem was, the company's supply chain didn't seem to recognize “make less this month than we did last month” as a viable plan. Instead, the planners kept following the system's advice to “make more.”

You can imagine what kind of havoc that played, not only on Cisco's system inventory but on that of its suppliers as well. Cisco had helped popularize the concept of contract manufacturing, where outsourced (or contract) suppliers built the routers and switches and then shipped them direct to Cisco's customers. Now, all of a sudden, Cisco's customers didn't want or need any more networking equipment—in fact, they already had too much. But Cisco's supply chain plan kept steadily insisting “make more.” The most important test of a supply chain plan is accuracy, and it became clear that Cisco was flunking that test.

A Bias Against Good Plans

Cisco's supply chain planning suffered from a common malady that afflicts many companies: bias. It's a pattern of behavior within a company where different departments focus on their own individual priorities, often disregarding the overall health of the company in favor of propping up their own silos. A good supply chain plan will fail every time, for instance, if employees are incentivized to avoid stock-outs and as a result keep building up the safety stock. Because employees are not being penalized for making too much—in some companies, the only unpardonable sin is to be caught short—the importance of the overall supply chain plan ends up taking a backseat to the size of one's weekly paycheck. When it comes to protecting and keeping their jobs, employees learned long ago that management will rarely punish those who tell it what it wants to hear.²

In Cisco's case, forecasting growth had been the right answer for more than 10 years, so it seemed the most natural thing in the world to keep going forward, even when it started to look like the boom days were over.

“There's a growth bias built into the business of forecasting,” explains venture capitalist Ajay Shah, a former director of Solectron, one of Cisco's major suppliers and one of the companies that got caught up in the undertow when too many unwanted electronics products started to flood the marketplace. “People see a shortage and intuitively they forecast higher.”³ That kind of growth bias leads to the unwritten rule of forecasting demand that says, “Err on the side of needing more, not less.”

Forecasts need to make sense, adds Si Gutierrez, senior director of supply chain with TDK InvenSense, a maker of motion and sound sensors, and a veteran of several other high-tech companies. A big part of forecasting for electronic device and chip manufacturers involves an analysis of general economic conditions. Gutierrez uses the cellphone industry as an example: “If the forecast says we'll need 20% more chips, we ask, ‘Does that make sense, given current market conditions?’ Everyone can agree that's a reasonable expectation for total market growth. The challenge comes in meeting with major players in the industry. Everyone wants to win and everyone's planning for success, so they add 30%. But not everyone wins. If you add up all the players in the industry, you might double a realistic forecast,” he explains.⁴

Ultimately, in the wake of the economic downturn in 2001, Cisco ended up with far more products than it could ever sell. How much more? The company wrote off $2.2 billion worth of unsalable, unusable inventory and reported a $2.6 billion quarterly loss. Although Cisco had gained the reputation of being the supply chain poster child for the New Economy, it reacted to the supply chain glitch in a typically Old Economy fashion: The company laid off 8,500 employees.⁵

From Soup to S&OP

How does a company overcome the inherent bias that seems to trip up even the best-laid plans? When Mike Mastroianni joined Campbell Soup, he saw many of the same cultural inhibitors to good forecasts that had stymied Cisco's planners. Brought in to oversee a sales and operations planning (S&OP) initiative at the world's leading soup maker, he found a supply chain that was focused too much on managing internal costs and not enough on customer service.

“For Campbell's, like a lot of companies, manufacturing was king,” remembers Mastroianni, formerly Campbell's vice president of planning and operations support and currently vice president of global supply chain planning with medical device manufacturer Medtronic. Manufacturing was in a position to second-guess the forecasts, thanks largely to the fact that some people had worked in that department for decades and had a historical perspective on how the market fluctuated. Mastroianni's mission, however, was to realign the supply chain to facilitate the introduction of new products. “Campbell's had become complacent,” he says, and to turn things around, forecast accuracy had to get a lot better.⁶

The average error rate of forecasts in the consumer packaged goods industry is 26%, but Campbell's wasn't going to get too far if it merely maintained the status quo. “We decided to focus in on forecast accuracy, which meant we had to change the behavior of bias,” Mastroianni explains. “People used to get their heads handed to them” for missing their numbers, so they tended to overforecast. As a result, they drove inventories up, as well as the costs of obsolescence, warehousing, expedited shipping, and everything else that was affected by overly optimistic forecasts.

How is a forecast created? No, they're not made up out of thin air, as some wags have observed. Campbell's, like many other companies, uses a traditional S&OP consensus process, which triangulates among sales, marketing, and demand planning. These three groups get together to agree on a number. That forecast number ultimately ends up going to the general manager for endorsement.

“Instead of aiming for a single demand figure, progressive companies have turned to forecasting a range of potential outcomes,” explains Yossi Sheffi, director of the MIT Center for Transportation & Logistics. “They estimate the likely range of future demand, and use the low end and high end to guide contracting terms and contingency plans.” The goal of this range forecasting is to get companies to widen their planning horizons.⁷

Even after consensus planning, though, the odds are pretty good that a company is not going to hit that number, which makes it all the more important that a system of open and ongoing dialogue is in place.

No Time Like the Real Time

One element driving Campbell's need for better forecasts is its collaborative planning, forecasting, and replenishment (CPFR) efforts with key retail customers. “We were forecasting at a very high level, based on history,” Mastroianni recalls, but to get to a truly collaborative relationship with its customers, the company had to be able to restate its history more frequently than once a month. Because CPFR requires manufacturers and retailers to share point-of-sale data over the Internet in real time, inaccurate forecasts only hasten the distillation of bad information.

“What fuels S&OP is facts,” he observes. That meant Campbell's needed to put key performance indicators (KPIs) in place to hold people accountable as well as measure improvements in forecast accuracy. Mastroianni's team turned to a real-time forecasting tool capable of creating daily, short-term forecasts with 52 weeks of live data. Being able to forecast in real time made it possible for Campbell's to track patterns that used to go undetected. The system might say, for instance, “Forget about the order today as it relates to your forecast. You need to be thinking about the next 7 to 14 days because, based on this current pattern, your next month is going to look like this,” he explains. “Or it might say, ‘You're holding on to a forecast that just isn't going to happen. So let it go, and produce to this lower number.’”

Campbell's learned, no matter how capable and experienced its planners are, their plan is only as good as the information that feeds it. The big “aha!” moment at Campbell's came when the S&OP process illustrated exactly how broken many of the company's processes were throughout the organization—from finance to commercialization to label design, custom pack planning, and transportation. S&OP provides a heightened level of transparency to the extent that, over time, as Mastroianni puts it, “the truth plays out.” By bringing all of Campbell's business plans into a single, integrated set of plans—the end game of an S&OP initiative—the company was ultimately able to fix a dozen or more major processes.

For instance, Campbell's was able to improve by as much as 50% the weekly accuracy of the item-level signals sent to its manufacturing plants, which resulted in an immediate benefit: The company was able to better plan how many trucks it needs to replenish its distribution centers with product. That increased level of accuracy also paid off by reducing how often Campbell's has to use expedited shipping to make up for not having the right products for its customers at the right time.

Taking it a step further, Campbell's leveraged its precision of accuracy to provide improved visibility to its warehouses and manufacturing plants. This has allowed the company to use its long-range planning capabilities to prebuy transportation with some of its carriers. Those forecasts have also been used for labor management, specifically to determine when Campbell's needs to add extra crews to its warehouses and when to cut back.⁸

Chemical manufacturer Dow Chemical uses a variation on S&OP known as executive sales and operations planning (ES&OP), which focuses on forecasting decisions made by senior-level managers that are grouped using an aggregate market-based process. This technique allows Dow to forecast from 3 to 36 months ahead, the difference being that ES&OP looks at overall market performance rather than drilling down to look at specific products or customers, explains Jacqueline Faseler, global director of supply chain sustainability. The idea, she says, is “to align the forecasting process with how the commercial and marketing organization thinks, whereas before we may have had businesses that tried to forecast what's going to be made on an asset or a product. We try to look more at market trends [than at] external factors.”

The payoff for Dow has been improved accuracy in its forecasts, which has helped its internal businesses focus on “bigger picture” market trends, Faseler says. “Businesses are not getting caught up in the details of what customer A, B, or C is doing next week or next month.” Instead, the ES&OP strategy “paints a higher-level picture, which forces the businesses to take a step back and analyze what's going on. They can make decisions around balancing demand and supply that maybe before they never got to because they were so focused on trying to collect very detailed data.”⁹

End-to-End Integration

One of the keys to an S&OP program is being able to integrate different departments and their various processes into one central plan, and that strategy can be applied in any company in any industry. At computer giant IBM, for instance, integration is not only a key best practice for the company, it's even included in the name of its supply chain organization, the Integrated Supply Chain. IBM's supply chain group comprises procurement, manufacturing, logistics, engineering, hardware operations, and sales support for all IBM products and services, overseeing $35 billion of supplier expenditures. The same group also oversees IBM's environmental initiatives.

Most of IBM's supply chain planning is done internally, involving such departments as logistics, fulfillment, manufacturing, and engineering, as well as functional experts in the company's business consulting and business transformation groups. Having all of these professionals in one organization allows IBM to tap into their expertise within each function, whenever needed, with that expertise fully integrated across the supply chain.¹⁰

Not surprisingly, technology has a lot to do with defining best practices within IBM's supply chain planning and forecasting processes. “Supply chain leaders are masters of disruption management, but they spend a tremendous amount of time fighting fires they never saw coming,” explains Jeanette Barlow, IBM's vice president of supply chain solutions. Digital technologies, especially artificial intelligence, can provide the kind of deep visibility companies need to build a smarter supply chain, the kind that can “proactively mitigate disruption and risk,” she says, “allowing them to focus more on the strategic needs of the business.”¹¹

Analyze This

As software has become more sophisticated and hardware more powerful, so too have performance tools evolved that allow companies to not only gain access to the supply chain-related data in their various systems, but to extract meaningful information that can provide them with an overall look at the current state of their operations. Or at least, they try. Supply chain planning systems are often integrated with enterprise resource planning (ERP) systems, which tie together manufacturing, sales, distribution, and finance by collecting data from each area and using it to plan a company's resource use—everything from employees to raw materials. All that planning, as you would imagine, involves a lot of data creation.

One popular meme cited frequently is that 90% of all the data that's ever been created was created within the past two years. It's been estimated that 2.5 quintillion bytes of data are produced every day, and that number is only going to rise at an even more accelerated rate as the data created by machines and shared through the Internet of Things (see Chapter 13) grows exponentially. It's fair to say that the era of Big Data has become the age of Humongous Data.

That being said, it's also fair to say that most of that raw data has limited to no value, at least until such time as something meaningful is done with it, and that's where the practice of data analytics comes in. “Data is the new currency of business,” states consultant Greg Siefkin, “but data all by itself isn't enough. To extract its value, business leaders need analytical tools and sophisticated algorithms to refine raw data into insights that can drive better business decision-making.” As he points out, supply chain operations historically have been numbers-driven and quite open to embracing meaningful metrics and analytical approaches to improving performance, such as the SCOR model (see Chapter 3). The methodical approach that tends to characterize supply chain professionals, Siefkin says, “creates an excellent foundation for cooperation between supply chain and data analytics professionals that can complement their respective skill sets by bringing together deep knowledge of business processes and advanced understanding of data modeling and analytics.”

Easier said than done, of course. Leveraging data and bringing together all those skill sets, Siefkin explains, requires several steps:

• Obtaining access to relevant data.
• Integrating that data from wherever it currently resides.
• Harmonizing that data to ensure its quality. “Data quality,” he says, “is a real problem for many enterprises,” and less than half of organizations fully trust their data to make important business decisions.
• Defining how the data will be used.
• Running the analytics and delivering the results in an insightful and usable format.¹²

Supply chain analytics are in an evolutionary stage, moving from the descriptive—where the data tells you what has already happened—to the predictive—where the data tells you what will happen given an analysis of similar past and likely future situations—to ultimately, the prescriptive—where the data tells you what you should do to achieve desirable outcomes. And thus, the tools available to supply chain professionals offer varying levels of capabilities, based on the needs of the user (and the price the user is prepared to pay for analytics expertise, which doesn't come cheap).

For instance, business intelligence software (which used to be known as decision support software) is similar in nature to artificial intelligence software in that both are rules-based, both focus on analytics, and both attempt to enhance human decision making. What business intelligence does, in a nutshell, is consolidate data from many operational areas of a company, analyze that data based on guidelines predefined by the company, and then present a dashboard-like overview of the company's performance. The latest generation of these tools offers a look not only within the four walls of a company but in fact within the entire supply chain of customers, suppliers, and third parties.

Even though companies have gotten quite good at collecting transactional data, many are still beginners when it comes to making sense of that data. That's where business intelligence technology offers a helping hand. Here are some examples of companies that have gained new insights into their operations from business intelligence solutions:

• How are airlines able to ensure all their flights leave and arrive on time, that every seat is filled with a paying customer, that all luggage is accounted for, and that everything runs safely? That's a tall order for any type of company, but with so many variables at play in the airline industry—weather, maintenance issues, limitations on service hours for flight personnel, air traffic patterns, to name just a few—it's a tall order to coordinate so much data in real-time. WestJet, one of Canada's largest air carriers, uses an analytics solution with visualization features and a dashboard presentation that provides as complete or as focused a look into the data as needed. For instance, airport managers can now receive situational dashboard reports that tell them at a glance exactly where every passenger's luggage is at any time after a plane has landed.
• If you ever wondered what happens to the data retailers collect whenever you use a loyalty card, consider the situation with greeting card giant Hallmark Cards, which runs 2,000 Hallmark Gold Crown stores and has point-of-sale purchase information from millions of card holders. So far, so good, but merely knowing who bought what, and when, doesn't always result in the right conclusions. Using business intelligence to create predictive models of consumer behavior, Hallmark discovered buying patterns that helped the company redirect its promotional efforts. For instance, instead of targeting once-a-year shoppers with Christmas offers, Hallmark found it was more effective to aim its online marketing efforts at customers who regularly visit its stores year-round. The company is now able to produce more customized direct marketing campaigns throughout the year, with better results.
• The US military relies on Lockheed Martin Missiles and Fire Control to develop and manufacture combat, missile, rocket, and space systems. Compiling and producing monthly financial reports was a tedious process that took two weeks to complete, taxing the resources of Lockheed Martin's IT staff. Part of the problem was that everybody with input into the monthly reports, from corporate executives to administrative assistants, had to wait for the IT department to produce the specific information that each individual required. After Lockheed Martin adopted a business intelligence solution, employees no longer had to go running to IT for everything. Now individual users can get the data they need whenever they need it, and in a format customized to their needs. Instead of taking two weeks to close the books, Lockheed Martin can now do it in two to four days.¹³

A Happy Ending

Improving its supply chain visibility has also proven to be the key to Cisco Systems' rebound from its forecasting nightmares, which were described at the beginning of the chapter. The company's turnaround began with a dramatic paring back of suppliers (from 1,300 down to 600) and the concurrent outsourcing of logistics, subassembly manufacturing, and materials management. All suppliers and distributors can now tap into the same supply chain network, and as a result everybody has access to the same forecasts and is working off the same demand assumptions.¹⁴

Not only does the supply chain network save Cisco millions of dollars by eliminating paper-based purchase orders and invoices, but it also has improved on-time shipment performance. And by applying “analytical rigor” to its supply chain plan, the company can make better decisions sooner in the process, such as what to do if a key supplier can't meet its commitments. By optimizing its supply chain plan, Cisco was able to remove emotions and bias from decision-making processes. Since then, the company has shifted its strategy from selling technology to selling business outcomes, a transition so complete that in 2020 Cisco was ranked as having the top supply chain in the world, according to analyst firm Gartner.¹⁵