Photo: @faaphoto via Twenty20
Will old-time retailing skills fix the supply chain mess created by COVID-19?
Once retailers restart their businesses following the COVID-19 outbreak, they will be faced with challenges for which old skills may provide faster answers than highly automated and mathematically intense solutions.
The outbreak has created a series of anomalies within the management of the supply chain. Let’s play out a few likely scenarios and look at why some touted solutions may not be up to the challenge.
On the demand side, recent sales history is completely skewed. For some essential categories, the numbers are off the charts. For others, sales numbers fall between zero and extremely depressed. The mix between online and stores is also well off of norms, a scenario that is likely to carry forward, perhaps to a lesser degree.
If your allocation works by using the last X weeks of history, chances are the next allocation is not going to be particularly useful. Typical replenishment systems can adapt to peaks and valleys, but recent sales tracking is beyond that scope.
Does your system recognize panic buying as a temporary phenomenon that will result in future diminished sales or, more likely, will it think that toilet paper sales uplifts are a real trend and increase orders just when demand drops off?Â
Even if your replenishment solution can be beaten into submission to produce realistic order projections, will it create orders for products unavailable in the time frame desired due to supply chain dislocations on a global basis? Constrained supply, substitutions and erratic delivery intervals are not friends of modern systems. While an “old fashioned merchant” might know how to manage these challenges, hoping that “set it and forget it” automated solutions will figure it out is not a strategy.
Now’s the time to teach your modern merchandising solutions some old tricks to be reactive at scale.
Restating history will be important for planning solutions. This doesn’t mean just replacing H1 2020 numbers with H1 2019 instead. Buying patterns are changing (even top to bottom ratios are different now), so product mix is likely to be different. The process will take thought, and each retailer will have to come up with the right logic for their own business.
Consumers are likely to be more accepting of substitutions if their favorite item is unavailable. Retailers may find that planning/replenishing/allocating at a category (or category attribute) level is a good strategy. While it may lack precision, it may just meet the customers’ needs.Â
Sales patterns going forward are a guesstimate away. Anyone who thinks they can predict what products consumers will buy — at what prices, where and in what depth — is a future retail Hall of Famer. Retailers might do better to build reactivity into their merchandise strategy. This might mean for example ignoring a demand plan and going to simpler replenishment strategies, such as model stocks or sales replacements. It may be better, albeit costlier, to replenish stores more frequently with smaller deliveries. Back up stock would be held back to replenish whatever stores or fulfillment centers need it most.
Maybe it’s time to bring old skills to new problems.
Discussion Questions
DISCUSSION QUESTIONS: Where do you see the biggest challenges for essential and non-essential retailers in planning, replenishment, allocation and pricing coming out of the coronavirus outbreak? Will retailers need to reduce their reliance on automated supply chain solutions in the immediate aftermath?
Poll
BrainTrust
Business knowledge and intuition are required more than ever. This needs to be married to smart data science and AI solutions. It is inconceivable to manually analyze every single SKU.
Businesses need to analyze their product mix and the sales trends during this time. Some products will experience pent-up demand and recover lost ground. For some demand will never come back. For others, demand will come back pretty quick. Some have experienced panic driven demand now, and will actually come down in time.
Similarly customer personas and affordability will have changed. Some geographies will have been affected more than others. In short, this will require a lot of critical thought and analysis, before execution.
I don’t think they will need to reduce reliance on their supply chain solutions because they can leverage the tools to smooth out the spike. They can then look at weighing year two and three data more heavily while also heavily weighing current state movement. Although not pandemic specific, most tools handle disasters like hurricanes, tsunamis, floods, etc. and can be tweaked to lengthen the timeframe of the anomaly. That being said, automated warehouses cannot deal with this pandemic and the insane stress placed on them to deliver three to 10 times the orders. It takes at least 12 months to build an automated facility so people are the only immediate solution to pick, pack and ship or deliver.
First, each retailer has to make a non-scientific decision and that is; for what time of the year do I plan to need to be back in business? That questioned is not answered easily. Automated systems have the capability to smooth out anomalies. This period (or periods) of restrained consumption will show up as an anomaly in most merchandise categories. The idea of more frequent purchase orders and shipments to points of sale offers the least amount of exposure to guessing incorrectly. This is not easy to accomplish and it is not likely to self correct in very short order.
What a great framework to help focus some solution-oriented thinking! What does AI, or automated anything, really “know” about a scenario that could never have been predicted, and therefore modeled? All of the sudden, what used to be “knowns” in the near and medium term are now “unknowns.” So it’s human intervention to the rescue.
When will demand normalize — first for essentials, and then for non-essentials? Hearing the explanation for the toilet paper phenomenon was highly educational. Different brands with different products with different supply chains for different customers. And almost zero interchangeability between the different supply chains and customers. It may take a year or more for predictability to return to anything resembling normal. And the “new normal” is going to be a moving target.
Being in the middle of this very thing, it has been like a circus — jumping through hoops to make adjustments to a legacy warehouse system to make it work and fill orders as best as possible while waiting for vendors to catch up with product delivery. I am fielding calls each day about orders only being half-filled or not filled at all.
It takes understanding on both sides to know this is not intentional and inventory levels and order replenishment will get back to some normality soon. That understanding needs to be communicated to the customers at stores. I think they generally understand – they just want to know.
These are unprecedented times and, certainly, a call to action to re-imagine the retail planning, replenishment, and allocation processes. Automation has its place, and even with the most optimal predictive analytics-based systems, machine learning would never be able to predict the phenomenon that we are facing today with the pandemic. Retail merchandising is all about the marriage of the arts and sciences, including leveraging intuition and experience along with all the valuable machine-based insights.
What will hopefully come out of this experience is a contingency-based merchandising supply chain that has a backup plan for when the supply chains are strained due to unprecedented demand. This would require the human touch, which is a blend of ingenuity and the analytical sciences. Perhaps long term near sourcing supply chain capabilities makes sense, especially in times of crisis, where longer lead times will really impact the product distribution timeline.
This has been the ultimate test for all retail and supply chain systems. One thing that is clear is that for forward planning and forecasting this is a period that will need to be excluded from history.
The bullwhip effect in terms of delayed signals between shelf and supply has not helped things. Reducing this lag would help automated systems. The starting point is an understanding of current inventory – something that is still illusive for so many retail organizations because the inventory information is spread across data silos. Addressing this first has to be the starting point for any automated supply chain solution.
Overall, I agree 1000 percent with Peter. Retail sense is superior to the science at this point. Nothing is comparable.
We have to split this by vertical, I think.
It’s time to be a real retailer again!
I hope all major retailers are using a combination of good merchants with excellent solutions to get the best results. If anyone thinks a black box with no input from a merchant will provide the best results for all items at all stores then they are mistaken. However, to go back to replenishment at category level or manually overriding modern solutions is a step back in time. The problems mentioned were issues in the past but a modern solution can work out what is a spike caused by exceptional circumstances and still provide a better result than a completely manual process. Modern solutions have the ability to analyze this sort of activity, then suggest the right strategy and order volumes. A really smart solution would already have taken the current situation into account as it unfolded, looked at building additional safety stock and understood the good and bad supply channels taking these into account as well. Let’s not head back into the dark ages just because we have nothing better to do.
To most of the grocery players I have spoken with, the answer is clear: “we’re going to have a big job in resetting the stores for this particular time, and that will have to quickly change when this crisis passes. The vendors will have a significant challenge and responsibility in resetting their product back into their footages, because it must be done quickly.”
The system, and therefore the systems that support it, are broken. All of the POS data generated from the point where the public first reacted to the last day of the pandemic will be useful — but only to model what happens in a panic. So year-over-year comparisons are, as Peter points out, going to be pretty useless to anyone but the crisis planers. But the danger going forward post-pandemic is from old thinking, not old systems. Peter, for example, argues that consumers will be more willing to substitute having experienced forced substitution. I could make an equally strong argument that the opposite is true and that deprived shoppers will demand the products they have been denied for months. The truth is, we simply don’t know. So while it’s comforting to thing things will get back to “normal,” the truth is nothing will ever be the same for the generations who collectively experienced the pandemic. The real question is, will those changes become permanent? The real answer is nobody (honest) knows.
I love this post. What we need to also add is the financial challenge of exiting this shutdown with enough cash and credit to resupply stores with what will be needed at that time.
We’ve lost roughly a season of selling and need to think carefully about what should be in our stores when they re-open.
It’s time to listen to timeless advice attributed to Albert Einstein: “We can’t solve problems by using the same kind of thinking we used when we created them.” The supply chain needs to re-emerge with a focus on safety (products, associates, and shoppers), speed to shelf, and sourcing.
Today’s situation highlights the importance of real-time data to know where items are at all times. By itself that is not enough. Someone or some program with alerts needs to be monitoring the data to identify within a few minutes when a drastic change happens, identify a trend or anomaly, and make immediate supply chain changes. It will also be essential to have supply chain partners that can respond quickly. No longer can the past be used to predict the future. We have now seen how quickly the future can change.
Prior to the coronavirus outbreak, even the most advanced supply chain solutions required the input and knowledge of seasoned retail professionals to ensure they were configured properly. And even though they are getting smarter and more automated, they are ultimately just tools to help merchants do their job.
Immediate plans and forecasts for the balance of 2020 will not be able to rely wholly on 2019 data nor on trends earlier in the year. The “new normal” will be a dynamic and unpredictable situation for most retailers for at least the next few quarters and will require savvy merchants to use their own insights alongside these advanced tools to predict and react as quickly as possible as trends slowly work their way back to something with traditional curves.
My first thought is that any supply chain that can’t adjust to changes in demand — even wild ones like we’ve seen, which while highly unusual aren’t unprecedented (think natural disaster) — isn’t so much “automated” as “mindless.” Yes, of course, a human should review whatever a computer comes up with (I would think they do anyway); and just as assuredly, this will tell us which companies have robust policies/procedures in place, and which just have automated rubber stamps.
Retail demand signals have been shredded by consumer panic buying in “essential” FMCG and the scary chill in fashion and most hardlines. There is a single cause for this — Covid-19 — but the consequences and fixes are very different.
For most department stores and apparel retailers, this will likely be a “lost season.” They will need to liquidate or write off inventory to free cash for the next open-to-buy. There is some potential for “revenge shopping” when the rebound comes, but the scope and timing will be very hard to predict.
Grocery stores face a dissimilar consequence — unprecedented store-level sell through that has galloped ahead of an otherwise high-functioning supply chain. They are in strong cash positions, but they can’t obtain sufficient quantities to replenish the shelves. Their ordering systems, which depend on finely-tuned forecasting and timely delivery, won’t function properly in the absence of a coherent demand signal.
Let’s be clear — massively atypical shopper behavior is a STORE problem with supply chain consequences, not the other way around. The present challenges are not evidence that existing planning and replenishment systems are suddenly inadequate. It would be fallacious to conclude that retailers must pour more cash into safety stock to somehow prevent other unforeseen disasters. No amount could possibly be enough, and the supply chain will catch up as home pantries run out of room.
Tip of my hat to Peter for launching this very important discussion.
Retailers shouldn’t be thinking that they must decide between humans or machines. It’s true that systems that simply rely on historical data are not going to perform well right now, but that doesn’t mean retailers should revert entirely back to human intuition alone. Rather, retailers should be looking at other data sources – other countries or regions, pandemic statistics, etc. – to start building new, advanced analytic models to better anticipate demand in the short term. And once they understand demand, retailers can make more informed buying, replenishment and allocation decisions, as well as leverage it to make pricing and markdown decisions to maximize margin and sell-through of existing inventory.