As I left the podium at a conference a few months back, someone approached me with an unusual question: “Why do you hate point-of-sale (POS) data?”
I replied, “Hate is a pretty strong word; let’s just say it’s not all it’s cracked up to be.”
I have always been a bit underwhelmed by POS data. Having used it for more than 20 years, mostly for descriptive analysis, it always seems to be lagged and not complete enough for real decision-making. So, I concluded that its best use was as a directional guide to consider with regard to the planning process.
But, with the advent of new technologies, demand signal repositories, better latency and predictive analytics, is it possible that POS data has the potential to become a primary demand signal? The notion that a supply chain can be forecast (and then planned) with POS data sounds appealing and logical in theory, but my experience suggests this idea will not hold up practically, and supply chain management professionals can’t afford to live in a theoretical world.
There is no doubt that both POS and shipment data have issues in terms of their potential use as a primary demand signal. I could easily take the fallback position that most businesses forecast shipments, not POS, because that is the demand signal they receive from retailers. This would be the path of least resistance. But I wanted to test the theory, so, in my early years here at Combe, I tried to determine the quality of each set of inputs. Here’s what I found.
The complaint most often cited about shipment data is that it is not a pure demand signal from the consumer. This is true, of course; shipments are the physical manifestation of an aggregation of requirements. They are the net result of a forecast, back orders, inventory target levels, on-shelf and at-retailer distribution center service level requirements, and promotional volumes. Included in every order is a complex entanglement of rules, heuristics and assumptions. And therein lies the problem: Shipment data is complex.
Shipments are influenced by a dizzying number of factors, yet those same factors must be assessed to determine the health of the replenishment process and the business itself. Unraveling what caused a retailer to send an order for an item in a certain quantity can be both perplexing and insightful. Asking why usually offers some real business insight. Interestingly, the opposite can be true with POS data.
Another common grievance is that POS data is too simplistic. It is nothing more than a pull-related signal, devoid of nuance or meaning. A POS transaction is a static measure of a buyer’s consumption at a precise moment in time — a single instance absent any of the innumerable, everyday interactions that occur on a shelf, with a consumer, in relation to promotions, about inventory, as part of a shipment or during delivery.
Consider promotions, for example. Actual shipments related to most promotions tend to occur well in advance of the promotion date. By necessity, this puts shipments and POS activity out of phase. In theory, we can forecast a POS lift and align a shipment plan accordingly, but this is more complex than just planning shipments with some anticipated promotions. So, why not do just that?
Some proponents suggest that discrete retailer POS data — such as retail link data provided by Walmart — may form the basis for better planning of base order flow and promotions. This appears to be a compelling argument. And while it may be true for Walmart, it doesn’t account for the impact a rollback may have on other retailers that are fighting for the same consumer. Their shipments likely will fall during a rollback period, affecting the balance of their businesses. Thus, the lesson is simple: The impact of promotions needs to be considered relative to the net effect across a business as a whole — not just at a specific retailer — because there is always blurring between the channels. Therefore, this argument dissolves under the pragmatic scrutiny of everyday planning.
Another Achilles heel is that POS data does not account for trade inventory movements, which come in many shapes and flavors. For example, a downward change in open-to-buy at a major retailer is likely to influence trade inventory by reducing shipments even while POS indicators remain high. Just a simple change in a retailer’s replenishment system — like increasing an in-stock shelf fill parameter from 98 to 99 percent — is likely to increase both trade inventory and shipment levels, while having minimal influence over POS. We see this often during high-foot-traffic times of the year. Even implementing new software can, and often does, disrupt shipments, while POS indicators remain steady.
There are other instances when external influences that affect shipments do not show up in POS data, such as when a retailer takes an inventory reduction at or close to year-end, influencing shipments, yet barely touching POS. In fact, trade inventory movements are so prevalent that many companies use a ship-share model to help net out the impact of trade inventory shifts on the demand plan.
Likewise, countless ordering dynamics influence shipments. Minimums for order size, pallet layers and unit buys all inflate shipments relative to actual consumption. But all are necessary factors to consider because they add volatility to the demand stream. Similarly, outlet growth — the actual escalation of the number of stores at retail — causes mini pipe fills, which has a material impact on shipments but would not otherwise be captured in POS data.
Finally, POS demand-signal cheerleaders will tell you that POS data generates better statistical forecasts. To which I say, “Of course POS is able to provide a better forecast; it does not have all of the other external factors burdening or influencing it. However, those very factors must be considered by planners in order to maintain control of their business.
So, truth be told, I don’t “hate” POS data, but I do recognize its inherent limitations. Therefore, I strive to use it strategically and with proper consideration within my company’s demand planning and sales and operations planning processes. I must focus on the demand signals retailers provide, not some theoretical ideal. And for this reason, in my opinion, the advantage goes to shipments.
Patrick Bower is senior director of global supply chain planning and Customer Relations for Combe. He is responsible for the company’s sales and operations planning process, order management, and third-party logistics management. Bower may be contacted at firstname.lastname@example.org.