Lead Time Optimization: Groundbreaking New Technology or just applied Total Value Management-based Decision Optimization in disguise?

For those of you familiar with the fashion industry, you are probably familiar with Supply Chain poster-child Zara , who, through the use of innovative process and technologies, designed a flexible supply chain that allows the company to take a garment from design through the manufacturing process to store shelves in 10 days – an almost unbelievable turn-around time in an industry that typically takes three or four months to execute the cycle.

One of the technologies Zara uses to accomplish this impressive feat, and stay profitable, is Lead-Time Optimization. Lead Time Optimization, succinctly defined as the optimized selection and management of inbound supply chains with respect to (highly) uncertain demand and long production lead times, attempts to optimize product line profits by attacking and minimizing the key drivers that lead to markdowns, stock outs, inventory turns, and reductions in working capital, all of which have a negative impact on profit. Recognizing that the industry is fraught with uncertain fluctuating demands that cannot be predicted with any significant degree of accuracy in advance, the lead time optimization products try to optimize around demand risk that results from the impact of variable demands on production schedules, capacity, shipping, and associated costs by assigning dollar values to risks, speed, and flexibility.

Lead Time Optimization is the attempt to determine an optimal award configuration which maximizes supply chain speed and flexibility with respect to both the fixed and calculated costs. It works best if you are willing and able to consider innovative supply relationships where you guarantee a minimum overall buy with respect to both supplier capacity utilization and raw materials, since then the suppliers are able to guarantee availability when you need it.

On the other hand, Total Value Management, which David discussed in Supply Management or Spend Management? and Total Value Management, is the pursuit of optimal award decisions with respect to direct, indirect, and impact costs and supply chain risk. A procurement professional employing TVM sourcing techniques is generally interested in the lowest cost buy that will mitigate the identified supply chain risks to minimal acceptable levels.

Furthermore, Decision Optimization is the technology they employ to find the lowest cost solution that meets their risk-mitigation constraints. Generally, the objective function is designed to minimize cost, not risk or supply chain flexibility.

So how could these two technologies possibly be related? Let’s start with Lead Time Optimization. Although its proponents, which include SupplyChainge, one of the pioneers, and Infosys, promote it as a profit optimization technology and proclaim that it represents the highest value obtainable in strategic sourcing as it is based on industry-tailored data models, industry-specific optimization technology, and business process improvement, it is really just advanced cost minimization, where the cost that is being minimized is not the should cost, landed cost, or even total cost of ownership (TCO), but the average potential cost when you consider unexpected costs that could occur due to variances in demand. Furthermore, when you consider that profit is just revenue minus expenses and that there is no way to really maximize revenue when you have no control over how many items are sold and the final prices actually obtained, the only way you can really maximize profit from a procurement point of view is to minimize expenses, which translates into minimizing overall purchasing related costs – no matter what.

However, this “no matter what” is tricky business when you cannot determine in advance precisely how big your buy is going to be and your savings opportunities typically derive from spend leverage. However, you do have a demand range which will generally have increasing probabilities as you approach the center of the range, and based on this you can determine a “book” capacity that is likely to minimize your overall spend in the worst case scenario and yet keep your overall spend low.

For example, let’s assume your range is 20,000 +/- 10,000 with a 90% chance of it being +/- 1,000, 80% chance of it being +/- 2,000 and 10% chance of it being +/- 9,000. Let’s also focus on shipping costs for simplicity and assume that you can ship at most 24,000 units on a truck, that you don’t get the FTL rate unless you ship at least 18,000 units, and that any units that aren’t booked (or will not fit on the truck) have to go air freight. Then we might have the following situation:

         Truck    Air       Rates  Probability     Cost
10,000  10,000      0   1.5 /   0       0.0001   15,000          
12,000  12,000      0   1.5 /   0       0.0200   18,000
14,000  14,000      0   1.5 /   0       0.0400   21,000
16,000  16,000      0   1.5 /   0       0.0600   24,000
18,000  18,000      0   1.0 /   0       0.0800   18,000
20,000  20,000      0   1.0 /   0       0.0900   20,000
22,000  22,000      0   1.0 /   0       0.0800   22,000
24,000  24,000      0   1.0 /   0       0.0600   24,000
26,000  24,000  2,000   1.0 / 2.0       0.0400   28,000
28,000  24,000  4,000   1.0 / 2.0       0.0200   32,000
30,000  24,000  6,000   1.0 / 2.0       0.0001   36,000

So what do you pre-book to minimize costs? Well, pre-booking 18,000 units gives you your lowest cost from a logistics and transportation perspective, but what if you demand turns out to be 27,000? Then you are paying 18,000 for the pre-booked units plus 18,000 for expedited air freight or 36,000 in total! At a simplified level, what lead time optimization would try to do in this situation is determine the booking level that minimizes the overall average logistics cost regardless of actual demand.

At a basic level, it is essentially trying to find the value such that:
Avg (booked demand cost + variable cost) over all possible demands is minimum (while balancing the required supply chain flexibility demands). Note that the variable cost also applies to lower demands if the provider bumps up your rate for not shipping sufficient volume (and wasting capacity). Even without regard to probabilities or flexibility constraints, this is not a simple calculation. To do this by hand, you would have to calculate the cost of each potential demand relative to each booked demand, take an awful lot of averages, and then choose the best price, which might not be obvious. For example, lets look at the key breakpoints of 18,000 and 24,000 and the median, 21,000.

Booked Actual Base Cost  Var Cost
18,000 10,000    18,000     4,000 
       14,000    18,000     2,000
       18,000    18,000         0
       22,000    18,000     4,000
       26,000    18,000    10,000
       30,000    18,000    30,000
21,000 10,000    21,000     5,500
       14,000    21,000     3,500
       18,000    21,000     1,500
       22,000    21,000     1,000
       26,000    21,000     7,000
       30,000    21,000    15,000
24,000 10,000    24,000    14,000
       14,000    24,000    10,000
       18,000    24,000     6,000
       22,000    24,000     2,000
       26,000    24,000     4,000
       30,000    24,000    12,000

Which is best? Is it the middle value, which is nearest the forecasted demand? Or the higher value, which would keep expensive air freight to the absolute minimum? It’s not an easy answer. It really depends on the skew of your probability distribution and the likelihood of demand significantly spiking, but note that it’s not 18,000, which our original table seemed to indicate. If the likelihood of demand spiking is higher than the likelihood of demand dropping, then a higher value is going to be better as variations will carry the minimal differential cost, and we will have a “safe range” of lower values where we will still get the quoted rate and run minimal losses.

Now lets look closely at Total Value Management, which is the optimization of total cost of ownership relative to supply risk and which constitutes an equal focus on spend management, to control costs and improve operations, and supply management, to mitigate risks and ensure supply. Unlike TCO, which focuses on direct landed costs and indirect usage costs, TVM also focuses on the impact costs of every decision. What’s the cost associated with the risk of dealing with a new supplier? What’s the cost associated with selecting a fixed capacity transportation option? What’s the cost associated with not locking up supply of a critical commodity or capacity from your most strategic supplier? These costs are quantified and considered when you use Total Value Management. The model that decision optimization is applied to contains your direct costs (PPU, Freight, etc.), your indirect costs (processing, waste, etc.), and your impact costs (supply risk, demand risk, etc) as well as your supply (chain) constraints. The generated solution obeys all of your supply chain restrictions, and if these are properly formulated, in addition to being cost effective, it will enforce a flexible supply chains as only suppliers that meet the appropriate criteria will be selected.

This leads me to ask: what’s the difference between Lead Time Optimization and Total Value Management Decision Optimization? Both technologies have essentially the same goals, the same best practices, and ultimately the same effects – controlled costs under risk, which ultimately leads to higher profits, and higher margins.

One Response to Lead Time Optimization: Groundbreaking New Technology or just applied Total Value Management-based Decision Optimization in disguise?

  1. I like this article, because it has always seemed to me that introducing more elaborate modeling carries the risk of moving the problem and not solving it. It’s certainly true that a simplistic model won’t account for important variances, and a more sophisticated model will. But the more sophisticated model requires the development of more sophisticated inputs, such as PDFs (probability distribution functions) and so on, that have as much uncertainty built into them as any other assumption one might make.

    One of the most useful outputs of sophisticated modeling is a user-friendly sensitivity analysis — in other words, the model identifies what specific variables or inputs cause the most dramatic swings in projected cost. It is then possible to identify the key “wagers” that are being made on the assumptions side, which in turn enables one to check on those assumptions to ensure that they were generated as carefully as possible, and aren’t just somebody’s careless guess.

    The second chart above is illustrative of this point.

Leave a Response