RSR Research: Cooperation vs. Coercion
By Brian Kilcourse
Through a special arrangement, what follows is an excerpt of a current article from Retail Paradox, RSR Research’s weekly analysis on emerging issues facing retailers, presented here for discussion.
What stands in the way of retailers and manufacturers truly integrating their inter-corporate processes? According to André Martin, co-author of Flowcasting
the Retail Supply Chain and founder and CEO of Factory2shelf, the biggest hurdle is a “deep misunderstanding about what truly drives the retail supply chain.”
Mr. Martin said the partners must first clear up these misunderstandings or they will continue to seek solutions that don’t apply.
“When people don’t understand a problem, they seek complicated solutions,” he said. “Once they understand the problem, they realize that the solution is really straightforward.” Mr. Martin points to the fact that even after decades and billions of dollars spent on complicated technologies; the retail ecosystem is still clogged with huge inventory levels and a persistently slow turn rate.
So where does it go wrong? According to Mr. Martin, the answer is twofold: the wrong information is often used in forecasting demand in the first place, and there was not enough visibility into the entire chain to take quick corrective action when it was needed. By “the wrong information,” Mr. Marin means “aggregations,” and specifically, aggregations of demand data one or more levels up from where consumer demand is serviced- in the selling environment. “Aggregating hides inaccuracies, and it takes away your ability to look,” he says.
Mr. Martin advises a different approach. “Don’t forecast what you can calculate.”
Here’s how he explains it: a retailer typically generates both store-level and DC-level forecasts; the manufacturing partner will produce a forecast for its distribution network and another for its manufacturing plants. All four of these forecasts are performed independently of each other, using different units of measure (eaches, case pack, pallet, etc.) and often different time frames. At every step of the way, a different result is generated, and those differences create “friction” (inaccuracies that should be – but often aren’t – resolved). Mr. Martin suggests that if the retailer shared true demand data at the transactional level, that data could be used to calculate appropriate quantities at each point upstream in the supply chain. The result would be both better order quality and improved service levels, while reducing inventory in the pipeline. And to the extent that the data is made available daily, all the partners co-managing the chain can react to changing conditions at the store level.
I asked Mr. Martin how Factory2Shelf’s approach differs from or augments the VICS CPFR (Collaborative Planning, Forecasting, and Replenishment) process. Mr. Martin sees the “Flowcasting” process as taking CPFR to the next level by refining the current collaborative planning concept.
He explains: “With CPFR, retailers share information about what they think they will sell with their manufacturing partners. The information is aggregated for all the stores supported by a retail DC. Manufacturers take on the responsibility to replenish the retail DC’s after the partners agree on service level and inventory turn objectives and on what each DC will require. This is an excellent start but, by aggregating to the DC level, the partners miss a key ingredient – what is happening on a store-by-store basis. Flowcasting uses unaggregated store-level demand data.”
Discussion Question: Do you also think true collaboration between retailers and manufacturers is plagued by a “deep misunderstanding” over the ultimate drivers of the supply chain? What do you think of Mr. Martin’s ideas around the sharing of demand data to improve planning at various levels of the supply chain? What hurdles do you see over implementing such a process?
I always thought that “what stands in the way of retailers and manufacturers truly integrating their inter-corporate processes” was getting the buy-in and wholehearted cooperation of the subordinates expected to implement the plans. In other words the worst plans, when implemented with enthusiasm and gusto, often outperform excellent plans with halfhearted execution. It’s up to the implementers to make plans work, not the planners.
We sometimes discuss the bygone days when huge food chains like A&P and Safeway operated many of their own productions plants. Wouldn’t they have had a heyday with today’s integrated supply chain programs? Total integration from production to retail, and everyone involved would have had a stake in its success. Smooth.
Today, several companies claim to be able to provide “cradle to grave” systems for retail product flow “if only.” “If only everyone in the supply chain would buy into our system.” “If only we had a reliable track record of integration with every type of system we encounter.” “If only our users would follow our directions precisely.” If only.
I have a deep-seated bias against even the most elegant retail solutions that overlook the human factor. That would be the implementers, not the planners.
I’ve been an admirer of Andre Martin for years. It is great to see him increase his exposure with his books and his software venture.
I agree that there is only one basis for forecasting no matter where you are in the supply chain. And that is store level demand.
In a retail industry where stock-outs and shortages are still a normal occurrence, it is vital that demand forecasts consider unfulfilled demand during periods where stock-out periods. This is tricky because often a consumer will purchase a different item during the outage. So companion items can show disproportionately high demand. Therefore, the “substitutability” factor of each item becomes an important parameter in the demand forecast.
The other key factors are promotion. If a competitive item is promoted, it will reduce the demand for the item being forecasted. Similarly, promotional spikes must be removed from full price demand.
Retailers should be showing their suppliers the aggregate demand forecast that stretch out for 12 months or more. These figures should also reveal the impact of stock-outs in lost sales and the influence of vendor promotions.
I have mentioned in the past in this forum (“Supply Chain Digest: Inventory Management – Are we Making Inventory Progress?” Feb 26, 2007) that forecasting is a fragmented process and there needs to be more dialogue between all the entities that generate demand forecasts for the same products. Working with both manufacturers and retailers, we have always integrated store-level data in our analytics and predictive processes.
But one more dimension not included in this great article is short-term demand generation due to the marketing activities of the manufacturer. Manufacturers spend a lot of resources in mass media communications to drive incremental demand for their products, which the retailer may or may not factor in their demand planning and this could drive significant forecast error.
No one talks about the “fourth dimension” when managing the supply chain. We talk about cube (length, width, and depth) for managing storage and load building but we don’t talk about time. Time is why we need forecasts and why we will always need different forecasts. POS data is great feedback and it should be used to highlight exceptions from forecast and adjust allocations, but to think that everyone can work from one forecast is naive.
The first forecast is a “capacity planning” forecast. Depending on the audience, this forecast may be needed a year ahead in order to prepare financial budgets or two weeks ahead to issue employee work schedules. In either case this forecast puts constraints on what can be done in the execution phase because it establishes a limit on available resources. In the case of the budget, it may be the decision not to open a new DC or store. In the case of the employee schedule it establishes the number of labor hours that are available.
As you move closer to the actual execution time, short term forecasts can be adjusted by recent results. Now you are no longer in a capacity planning role, but rather an “allocation forecasting” role. This is where the DC and flexibility in order quantities becomes important. Based on actual performance, the merchandise in the DC can be reallocated to the stores. Hopefully the capacity planning was right and there is enough capacity to meet demand even though the specific allocation is different.
Managing the supply chain is a little like guiding a mission to Mars with a 14 minute delay between communications messages (or whatever the actual time to Mars is). The more merchandise you can “cache” along the route the fewer out of stocks you will have down stream. Looking at this from a purely economical point of view, certain items may be important enough to justify the additional inventory and others will not. We often call this the “desired service level” and combine it with demand variance and magnitude of actual movement to determine safety stock requirements.
The fundamental starting point is to accept that no forecast will be perfect. As it has always been, the companies with good operators who can react quickly and appropriately to exceptions are the ones who will profit.
Of course, using the same data is going to increase the likelihood of better demand estimates for manufacturers. However, this assumes manufacturers are producing product to demand levels. Mark is right–this will work for books just fine. Those of us on the CPG side, however, are often working towards a corporate sales objective that has little or nothing to do with consumer demand. The CEO says 5% growth–all businesses have to grow by at least 5%, whether demand is there or not.
Andre Martin’s description of dysfunctional supply chain forecasting helps point the way to substantial cost reductions. However, some retailers and suppliers DO use a streamlined flowcasting approach, driven by end-point demand versus demand at every interim step. Over 600 retail chains with 140,000 locations report POS data to NPD, for example, and smart suppliers analyze this information carefully. In the book business, publishers get POS data from bookstores before determining reprint run quantities, because unsold books get returned to the publishers.