A Reflection on David Simchi-Levi’s Work
Among supply chain professionals, I am often asked how my views relate to those of David Simchi-Levi, whose textbooks and research have shaped much of the modern vocabulary of the field. It is a natural question: many practitioners learned supply chain through his models and case studies long before they encounter my own work. Our conclusions frequently rhyme, but the paths we take to reach them differ in important ways, and those differences — in how we frame the discipline, the future, and the role of software — have practical consequences for how companies design and operate their supply chains.
I have laid out my views at greater length elsewhere, most recently in my book Introduction to Supply Chain and in the essay Supply Chain as Economic Bets in a Market-Driven World, but my intent here is modest: to clarify my own perspective by placing it side by side with Simchi-Levi’s. I will focus on his widely read textbook Designing and Managing the Supply Chain and his management book Operations Rules, as well as his work on supply chain risk and digitization.
What are we actually managing?
If you ask a typical operations textbook what supply chain management is, you are likely to read some variation of the following: the integration of suppliers, factories, warehouses, and stores so that the right product is delivered to the right place at the right time, at minimum total cost, subject to service requirements. Simchi‑Levi’s textbook stays close to this spirit, and does a good job of developing models around it.
This definition is not wrong; it is simply incomplete. It describes the plumbing but not the game being played.
To me, supply chain is first and foremost an economic activity. Companies are not in the business of “minimizing logistics costs” or “maximizing service levels” in the abstract. They are in the business of allocating scarce resources that have alternative uses – cash, capacity, shelf space, even the attention of planners – in the hope of getting more money back, sooner rather than later. In that sense, supply chain is applied economics under uncertainty, instantiated through software. I develop this view more systematically – describing supply chain as a portfolio of economic bets under uncertainty – in Supply Chain as Economic Bets in a Market-Driven World.
Once you look at supply chain this way, the central object is not the truck or the warehouse, but the option: the ability to do something valuable when circumstances turn out one way rather than another. Inventory is an option to sell. Excess production capacity is an option to react to a surge. A second supplier is an option to avoid being held hostage. The point of the discipline is to cultivate and exercise those options in ways that generate superior returns.
Simchi‑Levi certainly acknowledges uncertainty, and he has written extensively on pooling risk, mitigating bullwhip effects, and designing flexible networks. Where we differ is that for him uncertainty is something to be controlled; for me it is the very substance from which supply chain extracts value.
Objective functions: money versus proxies
A second area where our lenses diverge is the question of what we are trying to optimize.
In Designing and Managing the Supply Chain, the standard formulation is to minimize total system cost – production, transportation, inventory, facility – subject to service level and capacity constraints. This is the language in which many optimization models in operations research have been written: a weighted sum of costs here, a service‑level constraint there.
In Operations Rules, Simchi‑Levi takes an important step forward. He insists that operations strategy must be anchored in the value proposition of the firm and argues that flexibility is the key enabler linking operations to customer value. This is a powerful message, and I agree wholeheartedly with the idea that flexibility is disproportionately valuable.
What I find problematic is leaving the objective function at the level of “costs and service levels” or “customer value” without forcing everything through the narrow gate of money and time. If we do not reduce our many metrics to something like a rate of return on the capital and risk we deploy, we are effectively optimizing proxies. Proxies are convenient, but they are also subtle sources of misalignment. It is entirely possible to improve service levels and reduce local costs while still destroying shareholder value once capital intensity, risk and opportunity cost are properly accounted for.
This is not an invitation to worship a single magical KPI. It is a call to recognize that, at the end of the day, supply chain decisions are investment decisions. They should be framed and judged as such.
Planning the future versus preparing for it
A third difference concerns our stance toward the future itself.
Simchi‑Levi’s body of work assumes, quite reasonably, that companies will plan. Forecasts are refined, push and pull processes are separated, inventory targets are set, and capacities are allocated. Better models, risk pooling, and contracts make these plans more robust. The underlying worldview is that there is a “good” plan for the network, and that our job is to approximate it despite uncertainty.
My experience in industry has made me wary of this planning mindset. Not because planning is useless, but because the way it is practiced tends to confuse prediction with control. Forecasts are treated as a kind of fragile truth about the future. Once agreed upon, they become a constraint to which everyone must conform. Deviations are seen as failures of execution rather than as signals from reality.
In my view, the future is not an engineering specification that we can meet if we only work hard enough on our planning spreadsheets. It is a contested, path‑dependent, and deeply uncertain environment shaped by competitors, regulators, customers, and random events. Against this backdrop, the key question is not “What is the right plan?” but “What portfolio of options do we need so that, when the future surprises us, we are more often helped than hurt?”
Simchi‑Levi’s work on risk exposure is very much about this portfolio, even if he uses different language. His Risk Exposure Index, built on the notions of time to recover and time to survive, provides a quantitative way to identify which facilities or suppliers pose the largest disruption risk and to prioritize mitigation. I applaud this. My critique is not of the tool, but of the residual belief that, once we have adjusted our network and run our stress tests, we are “back on track” toward a plan.
From my side, I advocate a style of decision‑making that is less about converging on a single forecast and more about continuously pricing and re‑pricing options as new information arrives. In practice, that means embracing probabilistic views of demand and supply, and injecting them directly into the logic of the decisions themselves rather than treating them as a separate forecasting activity that feeds planning meetings. I first laid out this critique of single-number forecasts and consensus planning in the “Forecasts, plans, and the illusion of certainty” section of Supply Chain as Economic Bets in a Market-Driven World.
The role of software: from records to decisions
Here, the contrast is less between Simchi‑Levi and myself, and more between two eras.
Simchi‑Levi writes in the tradition of operations research. His books present models – for network design, inventory, contracts, flexibility – and case studies that show how firms can use these models to improve performance. Information technology appears as an enabler: a way to implement better planning, gather better data, and, more recently, apply analytics and machine learning at scale. His more recent work explicitly combines digitization, analytics and automation as the three pillars of a modern supply chain.
I share this enthusiasm for data and analytics, but I place much more emphasis on software architecture itself. Over the past two decades, most large companies have ended up with a stack of “systems of records” – ERP, WMS, TMS and the like – to capture transactions. On top of these, they have built or bought reporting and planning tools that slice and dice history and help coordinate human decisions. What is still mostly missing is a dedicated layer whose sole purpose is to make and execute decisions automatically, at scale, under uncertainty.
These decision systems are not just “clever reports.” They encode actual economic logic: what is an acceptable trade‑off between the risk of stock‑out and the opportunity cost of capital; when it is rational to pay for a second source; how to reallocate scarce capacity when demand spikes in one region and collapses in another. They run daily, or hourly, with minimal human intervention. Their performance is measured not by projected savings in a business case but by realized cash flows.
Nothing in Simchi‑Levi’s writing contradicts this vision. On the contrary, his insistence that companies should use analytics and machine learning to drive better pricing, promotions and operations is entirely aligned with it. Where I push harder is in arguing that, unless companies treat the engineering of these decision systems as a first‑class software problem, they will never fully benefit from the models and insights that academics like him have been producing for decades.
Incentives and the human system around the models
Models and software do not live in a vacuum. They sit inside organizations populated by people with incentives, fears, and career plans. Here again, my focus is slightly different from Simchi‑Levi’s.
His textbooks acknowledge misaligned objectives across procurement, manufacturing, logistics and sales, and he discusses contractual mechanisms and coordination schemes to align supply chain partners. This is an important topic, and his work on contracts, flexibility and risk sharing is widely cited.
My experience with large implementations has made me more pessimistic about how easily these misalignments can be fixed by good will and clever contracts alone. Some conflicts of interest are structural. A software vendor paid per seat has little incentive to automate the work of planners away. A consultant who bills by the day is unlikely to recommend the simplest solution that would make his presence unnecessary. A function whose prestige depends on headcount will instinctively resist automation.
These are not moral failings; they are just predictable behaviors under certain incentive structures. For supply chain leaders, the implication is that architecting the decision system includes architecting the human system around it: who owns which decisions, who is rewarded for what, who has the authority to switch from manual to automated modes. Without this, even the best analytic frameworks will be tamed into something politically acceptable but economically mediocre.
Convergences worth preserving
So far I have emphasized differences, because they are what clarify my own position. It is equally important to acknowledge where Simchi‑Levi and I converge, because those convergences tell us something about the direction in which the field is moving almost regardless of philosophical starting point.
We both treat supply chains as systems and reject local, siloed optimization. We both consider uncertainty and variability to be central, not peripheral nuisances. We both think that flexibility – whether you call it flexibility or optionality – is disproportionately valuable, and that companies should be willing to pay for it. We both see data, analytics and automation as essential to any serious attempt at improvement.
These shared convictions are not trivial. Twenty years ago, the dominant narrative in many boardrooms was still about lean, just‑in‑time, and single‑sourcing – efficiency as an end in itself. Today, after lockdowns, wars and financial crises, the conversation is slowly shifting toward resilience, optionality and digital capabilities. Simchi‑Levi’s risk exposure work has helped push this shift into boardrooms and even into public policy circles. I see my own work as part of the same movement, though with a different emphasis and vocabulary.
Why the philosophical details matter
One might ask: if the prescriptions often rhyme – more flexibility, better analytics, more holistic design – why bother with philosophical disputes about objectives and the nature of the future?
Because, in practice, these philosophical details leak into design decisions.
If you think in terms of “minimizing cost for a given service level,” you will be tempted to treat service targets as exogenous and to solidify them into constraints. If you think in terms of “maximizing economic return under uncertainty,” you are more likely to question whether the service targets themselves are economically justified, and to adjust them dynamically as conditions change.
If you see the future as something that can be approximated by a single plan, you will invest heavily in planning cycles, meetings and consensus building. If you see the future as a source of surprises to be exploited, you will invest more in data pipelines, automated decision engines, and options that give you room to maneuver when plans inevitably break.
If you view IT mainly as a support function to implement better planning, you will buy another module for your ERP. If you view it as the medium in which your economic logic lives, you will worry about the separation between systems of record and systems of decision, about versioning of models, about experimentation and safe rollback.
Simchi‑Levi’s work nudges companies in the right directions on many of these fronts. My own contribution is to argue that we must go further: to treat supply chain as an economic decision‑engineering discipline whose natural home is software; to measure our success ultimately in money and time; and to build organizations and systems that treat uncertainty not as an enemy to be suppressed but as the raw material of profit.
On those points, the contrast is not personal. It is a choice every supply chain leader has to make.
