A Modern Tale of Long (Supply Chain) Tails -- Part I




As a little kid growing up in former (and erstwhile happy) Yugoslavia and watching my elders, day in, day out, downing dozens of strong Turkish coffees with their neighbors and relatives (while discussing sports, weather, world politics, and the neighborhood gossip) I would sometimes naively ask for a sip of coffee. The deterring line (a bogey-man tale) from my folks would be that “kids that drink coffee end up with a tail on their rear side.”

A few decades later (being currently admittedly addicted to Starbucks triple-shot espresso drinks), it appears that modern supply chains suffer from long tails, albeit not due to anyone’s premature coffee consumption. That (and much more) was the enlightening conclusion of the recent Webcast entitled “Long Tails and Optimizing Inventories” conducted jointly by AMR Research, ToolsGroup, and Supply Chain Digest.

The recorded Webcast, which took place a few months ago and was reportedly well attended (several hundred listeners) can still be replayed here.  There is also a related elaborate white paper from ToolsGroup entitled “Mastering the Long Tail of Demand.”

According to Lora Cecere, a Research Director of Consumer Products at AMR Research, the current “anatomic” supply chain business problem stems from the fact that products are continually proliferating. For instance, there has been a 15 percent increase in consumer products over the past three years.

Thus, over 16 percent of orders in the consumer products sector have a "stock-out" (whereby the customer is turned down and possibly lost in the long run). Most of these stock-outs are associated with a new product introduction (NPI) or promotional campaigns.

In addition to proliferating products, other trends that are driving companies to long tails are the following: the need for more frequent replenishment control (and thus more granular forecasting), adding replenishment locations closer to the end customer, and trying to deliver 98 to 99 percent service levels.

As an illustration, to go back to the coffee theme from the beginning of the post, imagine Starbucks Coffee or Peet’s Coffee & Tea companies trying to handle all of the possible coffee beans and tea flavors (beside other short-lived and seasonal merchandize) across the multi-echelon supply chains (central and regional warehouses) and hundreds of retail outlets. Indeed, what are the chances of over-stocking some and under-stocking the other stock keeping units (SKUs)? Quite large, indeed.

To make things worse, coffee retailers would not be the worst example of long tails, since the above-mentioned Webcast has pointed out some other businesses (like automotive aftermarket part distributors) that attribute even 98 percent of all SKUs to slow-moving items that contribute to over 60 percent of revenues.

Yet, 82 percent of organizations still measure the weighted mean absolute percentage error (WMAPE), which is an appropriate key performance indicator (KPI) for a small number of items with high volumes and high demand forecast predictability (that form the “head” of the supply chain). But this metrics largely masks the forecast accuracy of slow-moving products that have low volumes and low demand forecast predictability (and that form the ever-longer “tail” of supply chains).

Where Does the Tail Start? 

To be more precise, from a supply chain perspective, the tail starts there where demand becomes lumpy, and this lumpiness is measured relative to the Replenishment Control Frequency (RCF). Increasing RCF lengthens the tail. Namely, when a company controls replenishment weekly, the tail starts at SKUs with 0.7 line-orders per week, whereas in case of controlling replenishment daily the tail starts at SKUs with 0.7 line-orders per day (the 0.7 line-orders per RCF means a 50 percent probability of zero demand, or half the chance that this particular SKU will not be sold in this time period).

Confused, overwhelmed, embarrassed and whatnot yet? I guess so, but please do not feel badly about your lack of  knowledge on the matter! In fact, only a few companies have processes in place to manage this product complexity. According to AMR Research, only 26 percent of supply chain companies engage in an important cost-to-serve analysis that calculates the profitability of products, customers and routes to market to give a fact-based focus for decision making -- on service mix and operational changes -- for each customer.

In a nutshell, slow-moving products are increasing and important (as new products and promotions) to growth strategies. Traditional advanced planning & scheduling (APS) and enterprise resource planning (ERP) deterministic (formulaic) approaches are not designed to properly forecast slow-moving products. Stochastic or probabilistic methods are much more appropriate here.

Not All Slow-Moving Products are Create Equal 

Last but not least, all slow-moving products are not created equal, which requires new levels of focus and management. Based on the products’ importance and cost-to-serve, AMR Research has created a slow-moving product framework with four differing strategies below:

  1. “Buffer” -- for products low in importance with low cost to serve;

  2. “Rationalize” -- for items low in importance and with high cost to serve;

  3. “Maximize” -- for products with high importance and low inventory carrying cost; and

  4. “Focus” -- for items with high importance and high cost to serve.


For instance, to deal with the "maximize" quadrant, companies should improve demand sensing and reduce demand latency to improve the organization’s time to respond. AMR recommends using deeper statistical modeling tools, like the new applications from John Galt, ILOG Logic-Tools , Optiant, SmartOps, Terra Technology , and certainly ToolsGroup for inventory optimization and warehouse replenishments.

I would dare to add here i2 Technologies, Manhattan Associates, Barloworld Optimus, JDA Software (including former Manugistics), Adexa, Invistics, Smart Software and Logility. In AMR's opinion, the solutions like Retail Solutions (formerly T3Ci) and True Demand for retail outlet sensing, and Market6  and Prescient to build inventory strategies from point-of-sale (POS) data in direct store delivery (DSD) environments, can also help.

Concurrently with these, companies should implement agility strategies, like postponement, flexible manufacturing strategies, and agile supply to improve the supply response. However, they should only build responsiveness when they have a firm foundation of reliability in place.

For recommendations on strategies how to deal with other classes of products, see the AMR Research’s alert article entitled “Of Long Tails and Supply Chains”.

Zooming Onto ToolsGroup

Before delving into what Joseph Shamir, ToolsGroup’s chief executive officer (CEO), had to say about the long tails topic, a little introduction of the demand-driven inventory optimization company would be in order.

Like most of its inventory optimization peers (see my earlier blog post on MCA Solutions), privately-held ToolsGroup, with North American headquarters in Cambridge, Massachusetts,  and with several offices in Europe and elsewhere, traces its roots back to the academic world. To be precise, it originated at the Massachusetts Institute of Technology (MIT) and Draper Laboratories, a world-renowned research & development (R&D) center.

It was there that Eugenio Cornacchia, now ToolsGroup’s Chief Scientist and co-founder, began to develop a new inventory modeling and optimization technology that became one of the earliest such systems in the world. In 1993, Cornacchia co-founded ToolsGroup with Joseph Shamir, with whom he had previously worked at ITP Group, and the two were among the first supply chain professionals to apply and commercially implement advanced analytics and “stochastic optimization” techniques to achieve more efficient and robust solutions to common inventory problems faced by manufacturers, distributors and wholesalers in a range of industries.

Today, ToolsGroup continues to provide companies with inventory optimization solutions to help them achieve their maximum possible levels of customer service with less global inventory. Its solution has been successfully deployed in 31 countries (also with the help of specialized partners) at over 130 companies around the globe. As a proof of its history of success, ToolsGroup points out that 95 percent of customer licenses ever purchased are still in operation today, under maintenance.

Although not sanctioned by SAP under its prestigious Industry Value Network (IVN) clique (where SmartOps and MCA Solutions rather bask in their glory for ordinary and spare parts inventory optimization, respectively), ToolsGroup can still tout its bolt-on capabilities and information technology (IT) compliance (agnosticism). It has over 30 implementations within SAP’s ERP install base (at some of the largest companies in the world), and the product has the “Powered by SAP NetWeaver” designation. As for the underlying technology, ToolsGroup is a Microsoft .NET Framework and SQL Server shop.

ToolsGroup’s Distribution Focus 

If there is something that does not particularly impress me about ToolsGroup, it would be its non-intuitive (if not even deceptive) name – the company is certainly not a chain of hardware stores. In other words, it does not offer knives or saws for surgically cutting off those long tails.

What I do like about the company, though, is its sharp focus: not trying to be all things to all people even in the narrow niche of inventory optimization. In fact, it is important to note that ToolsGroup is that rare company providing inventory optimization focused mainly on distribution environments (that handle finished goods and maybe some limited final assembly operations).

The vendor also boasts nearly 200 live implementation sites (almost equitably in Europe and North America) – in stark contrast to its competitors, who only have a fraction of that. It is important to realize that inventory optimization for manufacturing and distribution environments are actually two different problems.

In the manufacturing part of the supply chain, the problem is a strategic one of final assembly  “postponement” – and it involves great scale and complexity, since the company must consider complex data such as bills-of-material (BOM) and routings. In order to make the problem manageable, several simplifying assumptions are required, as evidenced by the simplified demand models used by such systems.

Looking at the problem from another angle, many inventory optimization software solutions were primarily designed to solve strategic supply chain design problems, whereas ToolsGroup’s focus is rather on tactical and operational issues. The strategic issues entail large combinatorial problems that require very high scalability and therefore also require dramatic simplifications. These strategic approaches either totally ignore or grossly simplify the structure of the demand distribution in order to solve the problem.

The optimization is often based on quite sophisticated optimization algorithms, but by over-simplifying the demand model these solutions are actually optimizing a hypothetical model that does not match the real world behavior with a sufficient degree of detail. This approach (somewhat resembling the classical approximation of "spherical cows" in physics) may work well for supply chain network design, when the main purpose is answering core questions such as “How many distribution centers (DCs) should we use?” or “Where should they be located?”

But for the daily grind questions like “What inventory investment vs. service level should we set per main product lines?”, which are normally considered typical sub-problems solved by classical network design applications, they may not find the right answers when dealing with distribution situations, once the goods leave the factory. These assumptions invalidate the model for use in distribution, since experience shows that delivering high service levels (in the upper 90’s percentile) requires a much stronger and more complete statistical model.

Part II of this three-part blog series will depict the ToolsGroup solutions that are geared for distribution environments with long tails. In the meantime, please send me your comments, opinions, wishes, etc.

I would certainly be interested in your experiences with this software category (if you are an existing user) or in your general interest to evaluate these solutions as prospective customers.
 
comments powered by Disqus