How is omnichannel complicating demand planning?
Demand planning “has become an enigma” for supply chain professionals, according to the sixth annual “State of the Retail Supply Chain” from the Retail Industry Leaders Association (RILA) and Auburn University’s Center for Supply Chain Innovation.
Broadly, the challenges include an ever-increasing number of SKUs, tremendous price pressures, demanding customers willing to switch loyalties in a moment, and a multitude of order fulfillment options.
In interviews, 24 senior executives from a wide range of top retailers stressed the need to better predict demand quantities and the origin of demand (region, store, and channel).
Seasonal variation has also emerged as a more challenging planning issue. With demand fluctuating between channels during peak holiday seasons, retailers find it harder to predict where to position inventory. Amazon Prime Day in July and other major promotional events also present new challenges.
The reduced planning horizon that comes with omnichannel retailing is leaving little time for retailers to adjust demand forecasts. Wrote the study’s authors, “To counter this pressure, demand planners are now being asked to provide forecasts with greater detail. As a result, identifying an appropriate level of demand aggregation is another primary planning challenge.”
Finally, the use of multiple shipping points, which now more commonly include stores, adds difficulties when retailers attempt to manage inventory levels and accuracy.
One result is that only 31 percent of respondents to an accompanying survey of 74 supply chain executives considered their e-commerce forecasting capabilities “excellent.” Sixty-one percent agreed that e-commerce “greatly complicates” demand planning activities. By comparison, 63 percent believed their organization does an “excellent job” forecasting in-store demand.
Retailers are addressing some of these issues by improving alignment between different functional units, greater “analyst finesse,” and taking a channel agnostic approach to inventory. The author’s wrote, “They must couple these plans with innovative store replenishment and delivery processes to respond to changing demand dynamics in this complex planning environment.”
Discussion Questions
DISCUSSION QUESTIONS:
How is omnichannel and e-commerce complicating the demand planning process for retailers? Do you see any solutions to improve forecasting needed for e-commerce demand?
Poll
BrainTrust
Ralph Jacobson
Global Retail & CPG Sales Strategist, IBM
Adrian Weidmann
Managing Director, StoreStream Metrics, LLC
It’s not easy to manage demand planning in an omnichannel world. One item that the article did not address is the seemingly endless proliferation of line extensions. Perhaps by cutting the number of SKUs in some categories, retailers can make life a bit easier for themselves and their customers. With fewer choices, it should be easier to plan and easier for consumers to shop.
I am a bit surprised at the conclusions. If anything e-commerce and omnichannel should help forecasting total demand. A retailer can service considerably more customers with concentrated inventory than it can by spreading it out among multiple stores. And if you give the stores the power to cover an OOS for a customer directly from a warehouse, inventory can also be reduced in the stores.
The complications relate to the fact that e-commerce extends the complexity of targeting via demand forecasting. To solve this many retailers are leveraging the big data approach. The “See-Now-Buy-Now” trend is forcing brands to either produce a large amount of inventory before receiving demand or block production capacities in order to be more responsive, these moves putting brands at risk of overstocking, waste and returns. More granular demand planning that incorporates more details on the target shoppers.
An age-old problem (demand planning) that is getting ever more complex with additional sales channels. I am seeing some innovators utilizing machine learning to take the place of simple history-based forecasting. The cognitive analytics required for machine learning encompasses gathering external data from myriad sources including social chatter, news, local events (sports, etc.), weather and other sources. This is no simple task. Human gut feel is less and less of an input to forecasting and computers now take this data and see how these sources interact with each other. What this does is help build a demand forecast that becomes more accurate over time. A major retailer saw some 60 percent improvement, which is really significant when we typically get excited about single-digit improvements in forecast accuracy. This is where I see solutions to this challenge headed.
From the shopper’s perspective, Amazon already opened Pandora’s Box and allowed shoppers to search globally and find what they wanted on their terms. Shoppers have been able to purchase and have those items delivered to their doorstep. That capability is extremely difficult, if not impossible, for any brick-and-mortar retailer to replicate. Shoppers want localized selections at their retailers. Retailers and their vendors trying to address all of these localized tastes and trends are at a distinct disadvantage to deliver the immediacy expected by these shoppers.
I just experienced a similar frustration yesterday. I have been looking for a very specific oil stain for some redwood deck balusters. I went to the manufacturers website and found that the Sherwin-Williams store near me carried the product. Since Sherwin-Williams was advertising a 40 percent off sale this weekend I was indeed surprised and delighted! (Unfortunately too soon!) I went to the store and asked the clerk for the product (I showed a photo of the empty can I had in the garage.) He had no idea what the product was. Here was his quote — “That happens a lot. Someone agreed to sell their product without telling us. It would mean we would need to do a special order.”
Thank you — goodbye! He never even offered to follow through or look into the matter. FAIL!
Sherwin-Williams had better figure this out! Never mind about forecasting. They need to simply deliver what they advertise.
As long as the consumer is at all fickle, retailers will not be able to precisely forecast demand. Even if the consumer becomes predictable, which is highly unlikely, the weather isn’t and neither are competitors. Omnichannel and e-commerce require a new demand planning model that captures additional data from sources like social media and mobile to truly understand what fulfillment choices shoppers will prefer under any series of conditions. If it’s Tuesday, it’s raining and I’m scheduled to take the dog to the vet in the afternoon, the model will tell the retailer to deliver the pet food to the vet so the shopper can pick it up at the appointment. That info gets loaded into an algorithm that predicts the next purchase and ones like it so the system gets better over time. It’s the “when, where and how to get which product to the shopper, then apply predictive analytics,” approach.
Demand planning is the assessment of future orders/sales (which are not the same thing) by product area and by origin of the business (online, location, etc.). Frankly this presents challenges that are largely similar to the demand planning regimen we are used to. Better forecasts, etc., remain the constant. The result of a good demand plan goes back into supply chain planning to procure the right product at the right time.
The real challenge is the recognition that there is a new requirement called “fulfillment planning.” Where in the delivery chain (online, single pool of inventory, set up and inventory localized by store, which DC) do I position the purchased/reserved inventory to its best advantage for the highest margin in the shortest time?
Fulfillment planning is the new paradigm and requires all kinds of flexibility, organization and systems,and it is not just about customer online orders, it includes shipping to stores for sales, BOPIS and ship-to-customer. It requires thinking differently and is fundamentally at odds with the old method which was to plan inventory and assortments tightly associated with planned sales.
Step one is to recognize that the purpose of demand planning is to manage the supply chain. Step two is to understand that there’s a whole new world called fulfillment planning and to think about how to manage that exercise, placing in the delivery channels the product that the supply chain procured.
Even before “omnichannel” (anywhere, anytime retail fulfillment) demand planning was a huge challenge for even the best operators in retail.
Based on the metrics, you can make the case that over the past 10-plus years retailers collectively have not improved inventory productivity or in-stocks, etc. Out-of-stocks remain at 8 percent on basics and 15 percent on promotional items, costing retailers billions of dollars in sales, not to mention brand damage.
Moving forward the “brand damage” sensibility (inconsistent shopper fulfillment in a 360 inventory omnichannel world) will probably do more to motivate retailers to invest in inventory forecasting and demand planning technology robust and scalable enough to get the job done.
The data-driven technology forecasting and 360 inventory demand fulfillment solutions are there for retailers. They just need to commit to them.
Not everyone is a great retail forecaster and can meet the demands of omnichannel and e-commerce marketing. The creation and use of a forecasting model to accurately measure e-commerce demand and inventory needs is still more an art than a science, and changing dynamics only makes it even more difficult in many categories. Some supply chains are better left alone when considering this (especially in fresh grocery) as a potential model to follow.
It’s a little dangerous to generalize about these issues across multiple retail formats, as this research attempts to do, in part because forecasting of seasonal fashion merchandise is a very different animal compared with the forecasting of perennial fast-moving consumer goods.
This is one of the greatest operational challenges faced by supercenter operators, like Walmart, Meijer and Target, which sell a variety of products with very different rates of turn and seasonal traits. Even items’ propensity to be ordered online can differ widely. Can a single store-level forecasting and supply chain process be expected to optimize across such disparate types of goods and shopping behaviors? Is attempting to unify forecasting, ordering, replenishment and upstream supply under a single system really such a good idea? Why do we never seem to question that assumption?
I’m frankly a little surprised that retailers still seem to fret about folding in demand from online sales. Where online orders are fulfilled in stores, daily depletion of items should not look different from items carried out by shoppers, as far as the reordering system is concerned. Where online orders are fulfilled in separate facilities, these should look like “stores” to the upstream supply chain (albeit with somewhat different selling statistics).
Of course folding in demand generated from e-commerce adds a dimension to upstream supply chain processes. Fortunately, it should be possible by now to derive insights from online sales that permit comparison of online and in-store behaviors and by extension, online versus in-store forecasts. For some (but not all) goods, it should make sense to combine both demand signals up the supply chain to optimize warehouse ordering.
Omnichannel corporate structuring programs are having far more issues due to core business executive overlapping and multiple application software with little or no file compatibility. The software issues need to become a high priority fix especially for purchasing planning and logistics. The need is for a single real time file, input and inquiry structure, what is more important than knowing where your money is? And inventory is your money. A uniform single application modulated to fit the company’s omnichannel needs would make it possible to streamline executive management. It will remove a lot of arguing and make it easy to address predictability and inventory location needs.
Like with most areas in ecommerce, I foresee an increase in machine learning solutions in the future, as Ralph at IBM mentioned. There will always be a need to improve upon past demand planning models and a self-learning engine will solve that challenge by improving itself over time – it’s a scalable long-term solution to invest in.
On the one hand e-commerce makes it easier to plan because it is shipping from a central location and you are not worrying about allocation and standing inventory in store locations. On the other hand, store demand planning is much harder with consumers jumping between different channels.
I think the solution isn’t more accurate forecasting, but finding ways to build better flexibility into the system to move inventory in/out of store locations and warehouses quicker to fulfill demand, and make it very easy for consumers to order delivery to home from store when there is out of stock. The one area I have not seen much improvement is ability to quickly take an online order in the store from an associate point of view and also compensate the store appropriately.
Planning and forecasting in an omnichannel world is very complex and requires sophisticated planning tools. Retailers need to take a unified commerce approach to demand planning. They need to develop real-time inventory visibility and access across their four channels (store, web, mobile and call center) to enable the ability to ship from anywhere to anywhere. The challenge has always been to display the proper mix at store level (size and color breakdowns) while supporting the other two to three channels. Do I allocate 80 percent to stores and 20 percent to the other channels? Do I lower that to 60 percent to stores and hold back 40 percent for the other two to three channels and replenishment for stores? Do I allocate everything in season to stores and supply the channels from store inventory?
Many retailer have not upgraded their planning systems for many years and were not designed to accommodate a single channel silo. The good news is that retailers realize the issue and 63 percent of the respondents in BRP’s Merchandise Planning Survey indicated that they plan to upgrade or replace their merchandise planning systems within two years.
Fortunately, there are technology tools available in the marketplace to support advanced planning that incorporates customer insights and trends, cross-channel integration, competitive information, and real-time data. However, technology is only a part of the solution — to truly offer customers a seamless experience across all channels requires alignment of the organization and processes to support the technology solution.
We need new math to respond to this demand … real-time analytics and techniques like “dynamic re-allocation” to reset in season. Enigma, conundrum, paradox or puzzle are all words that describe a true challenge for today’s multi-channel retailers.
I love this! I know the article is old, but it’s actually more important today then when it was written. Now I’m a little biased being in GeoSpatial Artificial Intelligence and in this subject, but it’s smack on point for the complexity … and the “finesse.” Crazy complicated stuff, but it is no longer the future, it is now … today.
And solutions are needed because you can hire 100 McKinsey analysts to pull all this together … and they’ll have it in 6 months, when it’s already too late and the competition has zipped past you.
I believe all the solutions and stories that other writers have entered here are valid and on point, but wow would I like to hear what they have to say today.
To me this is “Demand Forecasting and Planning + Unified Commerce.” Okay, my head’s spinning just from typing that, but that is the holy grail (for this year at least). Or so I think. But love it.
Thank you Tom for the write.