CSCMP's Supply Chain Quarterly
October 17, 2019

Why has CPFR failed to scale?

The consumer goods industry has paid too much attention to planning and forecasting, and too little to synchronizing product flow with demand, according to a longtime industry observer.

When efficient consumer response (ECR) was introduced in 1993, supporters predicted that the demand-driven replenishment system would generate $30 billion of savings for companies in the grocery industry. Ever since, trading partners in the grocery supply channel have attempted to focus on the processes that would make ECR a reality.

Because so many organizations were quick to draw up blueprints for ECR implementation, many people joked that ECR stood for "every company's re-engineering" and that it contributed to "every consultant's retirement." But what ECR really should have stood for was "effective channel relationships," whereby supply chain partners work together to achieve the billions of dollars in savings promised by the initial efficient consumer response analysis.

Article Figures
[Figure 1] The problem: Dis-integrated systems across the supply chain
[Figure 1] The problem: Dis-integrated systems across the supply chain Enlarge this image
[Figure 2] The CPFR business requirement: Synchronous demand management
[Figure 2] The CPFR business requirement: Synchronous demand management Enlarge this image

Continuous replenishment programs (CRP), which trigger manufacturing and replenishment when products are purchased by an end user, and vendor-managed inventory (VMI), where retailers make suppliers responsible for determining order size and timing, were frequently cited as key processes in the quest for ECR implementation. But CRP and VMI implementations were rife with challenges. For one thing, they required the management of new and more abundant sources of information. They also required an understanding of the unique "rules of engagement" associated with each new relationship that was established among trading partners.

Maintaining competitiveness in a new, collaborative environment requires substantial changes in relationships along with investments in strategy, information technology, and infrastructure—on a physical, an organizational, and a cultural level. Moreover, synchronizing business processes with consumer demand—one of the principal objectives of collaborative programs—requires an unprecedented level of cooperation and communication among trading partners within the channel, with enterprises throughout the channel, and across enterprise functions. However, communication in a collaborative environment must be technology-enabled if it is to be effective. That's because manual communication methods simply don't "scale." In other words, they can't be efficiently and uniformly extended to all parties in a supply channel, regardless of their size, sophistication, or role in the supply chain.

That has proved to be true in the case of collaborative planning, forecasting, and replenishment (CPFR). In a CPFR program, trading partners jointly plan and manage supply chain activities, business planning, sales forecasting, and all other operations required to replenish raw materials and finished goods. (CPFR is a trademark of VICS, the Voluntary Interindustry Commerce Solutions Association.)

Since publication of its first set of guidelines in 1998, CPFR has held out the promise of dramatic reductions in inventory and costs across the supply chain. But according to CSCMP's "17th Annual State of Logistics Report," nearly 10 years after CPFR's initial adoption, supply chain inventory and costs have reached all-time highs! Clearly, ECR, CPFR, and other initiatives whose objective is inventory and cost reduction through collaboration are not yielding the expected benefits.

Despite its limited proven success, collaboration remains top of mind in company boardrooms. Indeed, Thomas Friedman mentions "collaboration" nearly 500 times in his best-selling book, The World is Flat. Yet for all of the interest and attention they have attracted, collaborative initiatives such as CPFR continue to "fail to scale." More than 300 companies have implemented some form of CPFR, but they represent only a small percentage of the companies that could benefit from this strategy. What's more, many companies have estab- lished strategic collaborative relationships with some of their most important customers and suppliers, but these implementations represent just a small part of their total customer and supplier base.

In this article, we look at some of the technological, organizational, and cultural constraints that have inhibited implementation of collaborative programs on a broader scale. Although collaborative initiatives have taken hold to some degree in automotive, high tech, apparel, pharmaceuticals, and other industries, our discussion will focus on the grocery and consumer goods industries, where many of these concepts were developed.

The impact of pricing and promotions
For quite some time now, the consumer goods industry has been undergoing a fundamental shift in attitudes concerning traditional business practices, particularly in relation to trade promotion and product replenishment. This shift crystallized with the formation of an industrywide working group and the issuance of a report in late 1992 that set the stage for the efficient consumer response movement.

In that report, the ECR committee proposed that the industry could save $30 billion annually and reduce systemwide inventories by more than 40 percent if more rational practices were adopted in four areas: trade promotion, replenishment, product assortment, and new product introductions. Actions related to replenishment accounted for more than 40 percent of the total benefits projected for ECR. Accordingly, replenishment has absorbed much, if not most, of the industry's attention on both the demand and the supply sides of the consumer goods channel, including retailers, manufacturers, and suppliers.

That said, the most important factor in these initiatives' failure to scale has been a lack of attention to how they are interrelated and to how cross-functional processes behave, separately and together, to affect supply chain inventory and cost. The result is that instead of being reduced, inventory and cost have been shifted elsewhere in the channel.

For example, manufacturers' pricing and promotion practices may encourage investment buying as well as diversion practices among retailers that result in inaccurate forecasts, inventory deployments, and production scheduling. New product introductions without consideration of existing product extensions and pipeline inventories contribute to rapid inventory obsolescence and create additional costs. Similarly, insufficient understanding of retailers' economics and strategies for product assortment contribute to forecast inaccuracies, higher inventories and costs for manufacturers, and inaccurate replenishment deployments. This lack of visibility is constraining scalable adoption.

For example, let's look at an early CPFR pilot, which involved a major manufacturer and a large grocery chain. During the initial meeting to set the pilot program's objectives, the manufacturer's vice president of supply chain suggested that the inventory objective for the retailer be set at 15 days of supply. The grocery chain's representative said that it couldn't be done.. Upon further review, it was evident that although the retailer's sell-through would support 15 days of supply, the retailer would actually have to carry 30 days' worth in order to maintain its high-volume "bracket one" pricing.

A gross margin return on investment (GMROI) analysis (the favored approach in retail economics) supported the retailer's decision to invest in an additional 15 days of supply to maintain the pricing/margin advantage from the discounts afforded in bracketone volume pricing rather than drop to bracket-two pricing. In fact, the retail chain could afford to invest in the additional inventory-turn cost because it was offset by the gross margin gain. Furthermore, there were several smaller retailers in the area that could not afford the inventory-investment cost required to take advantage of bracket-one volume pricing. These smaller retailers were more than happy to buy the excess inventory from the retail chain at a price that split the difference between bracket-one and bracket-two pricing (a practice known as "diversion").

The failure to consider the impact of pricing policy on replenishment strategy prevented the CPFR pilot from immediately moving forward. (This example related only to volume bracket pricing; think of the dynamic that could result from promotional pricing initiatives!)

But there is a second lesson to be learned here: Collaboration is not linear. Competitors often can collaborate to their mutual advantage. Knowing when to collaborate and when to compete can result in industrywide gains as well as individual gains. Collaboration has been common in the retail industry for many years. Not only do retailers work together on diversion, but many also collaborate in warehousing, transportation, private- label manufacturing/sourcing, and other non-shelf processes. Manufacturers, on the other hand, generally continue to consider end-to-end process control as an extension of brand competition.

The combination of a lack of visibility and insight into channel behavior and the failure to take a holistic view of account pricing, promotion, assortment, and new product introductions is a major contributing factor to the inability to scale and realize the promise of collaborative initiatives. To overcome that, companies must initiate a fundamental shift in the way they go to market and how they manage operations. This requires more than a focus on business processes. Instead, companies must align their market-development, decision making, and businessoperations processes. Furthermore, this alignment must be informed by an understanding of how their customers, suppliers, and competitors are carrying out those same processes. Using process-reference models such as the Supply-Chain Council's Supply Chain Operations Reference (SCOR) model can facilitate recognition of how processes across the enterprise and the channel interrelate and affect both channel and individual company performance. (SCOR is a trademark of the Supply-Chain Council.)

Rethinking POS data
In today's "flat world," converging technologies such as radio frequency identification (RFID), Web/Internet portals, enterprise business systems, business analytics and intelligence, infinite bandwidth, and wireless communication form a web rather than a straight line. As a result, they make the transformation from traditional, linear supply chain thinking to collaborative supply network thinking a competitive imperative. That holds true not only for channel partners but also for channel competitors.

Since 1973, when the universal product code (UPC) standard was first established in the United States, there has been an almost talismanic belief that technology would improve channel visibility, inventory velocity, cost reduction, and consumer responsiveness. More than 30 years later, that vision remains largely unfulfilled. There are a number of reasons for this. Among the most important has been the lack of consideration of and visibility to the demand spikes created by short-term promotions, new product introductions, assortment adjustments, competitors' initiatives, and demographic differences. But supply chain participants themselves have impaired the vision by their own inability and unwillingness to cooperate in support of open communication of information.

At issue is not so much the sharing of point-of-sale (POS) data, where most attention and investment has been paid. As illustrated in Figure 1, the bigger concern is that the retailers have not shared with the manufacturers how they use POS data to develop the forecasts and orders that they send to the manufacturer.

The enormous volume of data, as well as the inconsistency of data across the various points-of-sale and purchase cycles, makes it impractical for even the most progressive trading partners to use that information for forecasting the product requirements that support the different replenishment, production, and sourcing cycles upstream.

Each of these business cycles operates at a different clock speed, with different competitive product mixes and demand characteristics. For example, Walgreens cannot tell Procter & Gamble that the manufacturer's forecast for Crest toothpaste that is based on POS data is wrong because Colgate is running a "two-for-one" promotion. Walgreens can only tell P&G that its forecast is wrong and that it should be adjusted in accordance with the retailer's own buy forecast for Crest.

Sales at consumption points are small, inconsistent, and highly dependent on local demographics and an infinite number of constantly changing causal factors. That's why it's not the point-of-sale data that's relevant—the critical information is the rate of sale for each product, aggregated to the stores' sourcing-location level. Knowing what the store will order is easy. For the most part, they will order the pack-out case quantity. The real challenge is determining, based on rate of sales, when the store will reorder.

André Martin, who pioneered the distribution resource planning (DRP) system, likes to call this "flowcasting." But even before he coined the term, Martin had always stated that the only point of uncertainty was at the point of sale and that all other requirements upstream can be calculated.

André is right! Manufacturers should stop trying to forecast what will be sold at each point of sale. Instead, they should use their customers' forecast to calculate their customers' product requirements based on what would allow those customers to replenish their stores in sync with the rate of consumption for the manufacturers' products. As illustrated by the toothpaste example above, instead of simply analyzing existing POS data, it's a good idea to get information about any initiatives and campaigns that the retailers' marketing, merchandising, and sales organizations are developing that will influence consumption and cause the POS data to change.

From CFAR to CPFR—and maybe back again
Another major constraint on more widespread adoption of CPFR dates back to the earliest days of this concept. When ideas about collaboration first began to circulate, one of the trailblazers in this area, Ted Rybeck of Benchmarking Partners, conceived of and proposed the concept of CFAR, or collaborative forecasting and replenishment. Later, when CFAR became a VICS industry initiative, the members of the initial CFAR committee insisted that the collaboration process begin with joint business planning rather than focusing entirely on developing shared forecast and replenishment requirements. Hence, CFAR became CPFR.

The committee members made this change to try to address the major roadblock to adoption at that time: the interrelationship of different, highly variable factors that cause market demand to change on a daily basis. The fundamental difference between CFAR and CPFR was that CFAR sought to synchronize product flow with customer demand, while CPFR sought to bring market supply in line with market demand (based on data at the POS level) while stimulating store sales through demand-shaping initiatives.

This shift in focus from responding to demand to shaping demand resulted in "paralysis through analysis." Organizations simply don't have the time or the capacity to manually take on or to automate, on an account-by-account basis, the many processes and activities required for CPFR. I've seen this firsthand: Out of all of the category managers, buyers, and merchandise managers I've worked with over the years, virtually none has had the time to conduct joint planning with any company other than the category captain! Furthermore, the nature of these planning processes and the competitive sensitivities they may provoke make automating them difficult if not impossible— or even illegal, should authorities consider certain types of collaboration to be anti-competitive.

When we look at today's CPFR framework, we can see that something is missing. What's lacking is a reasonably straightforward capability to systematize the process of balancing and synchronizing retail replenishment requirements with manufacturers' shipping and production requirements. As a result, CPFR implementations have been relegated to the strategic few— that is, the largest customer/supplier relationships— rather than scaling to encompass the entire market. Unless the focus turns to systematizing and sharing the retailers' forecasts of the product requirements needed to meet store-level replenishment needs, CPFR will not be more widely adopted and deployed.

Thus, Sherman's Law of Forecast Accuracy states: Forecast accuracy improves in direct correlation to its distance from usefulness. Forecast accuracy at the manufacturer's shipping dock, by product and by customer load, is 50 percent at best. How much inefficiency, cost, inventory, and effort must be expended to compensate for the gap in accuracy between forecast and order requirements? Forecasts that are determined through statistical analysis of historical information often produce ridiculous results! No matter what the product, every day will differ in some way from that same calendar date in the past. After all, sales and marketing organizations, not to mention competitors and consumers, are dedicated to making the future different than the past.

Perhaps the most promising aspect of CPFR has been the growing recognition by retail and manufacturing participants of the need for more cooperative information exchange. Still, the open sharing of decision-support information (as opposed to transactional data) between retailers and suppliers has not occurred at the desired rate. As a result, even though many trading partners have initiated collaborative efforts, they have not realized the full range of benefits due to a lack of visibility into a "critical mass" of participants' processes and requirements.

Where do we go now?
By focusing on the flow of supply to consumers, channel partners can discover previously hidden bottlenecks and address them. Remember André Martin's premise that the only point of uncertainty occurs at the point of sale? All other requirements for "flowing" product in response to POS demand can be calculated and adjusted for that uncertainty. The problem is that the "dis-integration" of systems across the supply chain has created a perceived need to forecast demand rather than calculate product requirements at the fracture points (the points of operational and/or organizational transfer of goods) in product-flow visibility.

"Why should we forecast what someone else already knows?" is a question that Don Bowersox of Michigan State University has asked over the years. Maybe we should have spent more time in his class. The problem with building inventory and production requirements based on a point-in-time forecast of demand at multiple points in the process is that the demand forecast is always wrong! The consequences of such errors are dramatic inventory swings and costs, as seminal works by Jay Forrester, Peter Senge, Hau Lee, and others regarding system dynamics and the "bullwhip" effect have clearly shown.

Whether they are developed at the point of sale or at other points in the supply network, forecasts are always wrong. The difference between POS-based and other types of forecasts is that a major gap in accuracy at the point of sale usually results in a short-term error of perhaps one case. But even a slight forecast inaccuracy at a high-volume manufacturing plant or distribution center (DC) can result in millions of cases produced or deployed in error. The lesson to be learned from Forrester, et al., is that we must eliminate any delays in informing the network of the gap between actual and plan at the point of sale and of its impact on supply points upstream.

By independently and sequentially forecasting instead of recalculating (based on new demand data) at each point of fracture in the channel flow, we exacerbate and perpetuate the forecast error that occurs at the point of sale. The cumulative effect of forecast error throughout the supply network can make life interesting. My forecast is wrong, therefore my distribution plan based on the forecast is wrong; therefore my production plan based on the distribution plan is wrong; therefore my materialrequirements plan based on the production plan is wrong; therefore my purchasing plan based on the material requirements plan is wrong. And to top it all off, I've probably implemented a multimillion-dollar, integrated enterprise resource planning (ERP) system that will propagate errors at the speed of light and cumulatively "bullwhip" my resources as well as those of my customers and my suppliers.

In my view, when we collaborate, we need to separate the processes that create demand from the processes that fulfill demand. We need to share information about how and when demand creation will have an effect on fulfillment. But we also need to keep in mind that we can't manage demand creation and fulfillment simultaneously. As illustrated in Figure 2, our efforts should focus on **ital{synchronizing} those processes, not integrating them. Let's take the "P" out of CPFR and go back to CFAR. By doing so, we will gain visibility—in real time and at all levels and cycles in the supply chain—of the inefficiencies in the product flow that result from demand variability.

CPFR will continue to be one of the most promising and exciting initiatives to come out of business circles in the past century. Companies must embrace and participate in the continually improving best practices that are evolving from VICS' CPFR initiative. There is no doubt that it can be the catalyst for market leaders to transform their organizations from traditional, linear supply chain thinking to collaborative supply networks. And it need not be limited to market leaders: Supported by sophisticated business analytics and intelligence, enterprisewide search and communication capabilities, and business-process workflow and portals—all enabled by the Internet's unlimited bandwidth—companies of all sizes can collaborate on a large scale at very low cost.

But for CPFR to realize its global promise and improve overall industry performance, it will have to return to its roots. That means separating the processes for managing and shaping demand from the processes for replenishing product based on actual demand. CPFR will also have to advance best practices and processes that support broad-scale collaboration and are based on standards for systematizing the electronic sharing of information that is needed to manage product flow in response to demand. Only then will CPFR scale to a broader base of customers and suppliers across the industry.

Richard J. Sherman is president of Gold & Domas Research.

Join the Discussion

After you comment, click Post. If you're not already logged in, you will be asked to log in or register.

Want more articles like this? Sign up for a free subscription to Supply Chain Executive Insight, a monthly e-newsletter that provides insights and commentary on supply chain trends and developments. Click here to subscribe.

We Want to Hear From You! We invite you to share your thoughts and opinions about this article by sending an e-mail to ?Subject=Letter to the Editor: Quarter 2007: Why has CPFR failed to scale?"> . We will publish selected readers' comments in future issues of CSCMP's Supply Chain Quarterly. Correspondence may be edited for clarity or for length.

Want more articles like this? Subscribe to CSCMP's Supply Chain Quarterly.