Time in Supply Chain: Why the Future Is Not a Mirror of the Past
For most of my career, I have been puzzled by one persistent feature of supply chain thinking: the way we treat time. In Introduction to Supply Chain, I argued that the mainstream vision quietly borrows its notion of time from physics and meteorology, and that this import is deeply misleading in business. The present essay is an attempt to restate that perspective more plainly and to confront it directly with what most companies still do today.
The thesis is simple enough: the future of a supply chain is not a statistical echo of its past, but the consequence of the commitments we make today. Yet nearly all of our tools, rituals, and metrics are built as if the opposite were true.
How we quietly imported a physicist’s time into supply chain
Look at how most organizations approach their demand. They start from history: sales per product, per location, per week or month. They feed these numbers into a forecasting engine, decorate the result with business judgment, and obtain future time series. These projected lines of numbers, one for each SKU and site, become the backbone of everything else: capacity plans, purchasing budgets, production schedules, inventory targets.
This way of proceeding did not fall from the sky. It mirrors how time is handled in many areas of natural science. A meteorologist will record the temperature in Paris every day and think in terms of a time series – an ordered sequence of measurements along a time axis. The process is viewed as something that stretches indefinitely into the past and into the future. Yesterday and tomorrow are of the same nature; we just happen to be standing at “today”.
In physics, many laws are close to time-reversal symmetric. If you filmed an idealized pendulum and played the movie backwards, the equations would still hold. Time, in that setting, is a neutral dimension. You can move forward or backward; the formalism does not care.
Supply chain, as taught and too often practiced, has adopted the same neutral axis. Time is a line. Demand is a curve along that line. We believe that if we model this curve properly, we can then “engineer” the future by devising and enforcing a plan that matches it. S&OP, master production scheduling, annual budgeting: all those rituals presuppose this underlying picture of time as a smooth, largely predictable extension of the past.
On paper, it looks reasonable. In practice, it is a very costly illusion.
The illusion of symmetry: when identical histories hide opposite futures
To see what is wrong, consider two businesses that exhibit exactly the same weekly sales for a given product.
In the first business, a thousand independent customers each buy roughly one unit per week. In the second business, there is a single key account buying roughly a thousand units per week. If you only look at weekly sales, the two situations are indistinguishable. The time series is identical.
Yet any practitioner will immediately recognize that the risks are not comparable. In the second case, the next purchase order depends on the mood, incentives, and politics of one buyer. One phone call, one lost contract, and the demand can collapse overnight. In the first case, demand can shift, but it will typically do so by drifting across many customers and many decisions; it is much harder for everything to vanish in a single stroke.
From the standpoint of inventory, service level, and commercial effort, those two businesses call for radically different strategies. Still, the time-series representation – our supposedly “scientific” handle on time – declares them to be the same history and quietly invites us to treat them as the same future.
A similar problem appears if we keep the same average sales but change the way customers buy. Imagine a store that sells ten units per week of a given item. If ten different shoppers each buy one unit, a modest buffer stock will do. If, instead, one or two customers regularly buy five or ten units at once, a very different level of inventory is required to avoid stockouts. Again, the weekly time series is identical. Again, the economic reality is not.
The point is not that time-series models are clumsy; it is that they are blind to the very structure that matters. They collapse distinct futures into the same line on a chart. In doing so, they encourage us to believe that the future is just “more of the same”, plus or minus a bit of random noise.
But business time is not like that.
The forecast-and-plan doctrine: causality turned upside down
If the only problem were some loss of nuance, time series would still be a useful approximation. Unfortunately, something more severe happens once those time series become the centerpiece of planning.
Most companies operate under an implicit doctrine that I like to summarize as: forecast, plan, and enforce.
First, we forecast demand for the next year, quarter, or season. Then we derive a set of plans: how much to buy, make, move, or hire in each period. Finally, we organize ourselves so that operations “adhere” to those plans. Performance is then measured in terms of forecast accuracy, plan compliance, capacity utilization, and similar indicators.
This way of working quietly assumes that the future demand we are forecasting is largely independent of the very decisions we are about to take. It treats the future as an external fact, like the weather, to be measured as best as we can and then accommodated.
Take a simple example: a fashion retailer planning the sales of a new backpack over the next twelve months. A typical planning cycle will ask for a monthly time series of expected sales and then derive production and purchasing volumes from that curve.
But what will actually shape those sales? The final price, the presence or absence of promotional campaigns, the number and attractiveness of competing items in the assortment, the visibility given to the backpack in stores and online, the level of service achieved on that product relative to others. All of these levers are still open at the moment the forecast is produced. In practice, they will be adjusted multiple times during the life of the product.
And yet, the act of forecasting encourages everyone to speak and think as if sales next May were already “there”, waiting to be served. We model demand as an extension of the past, and only then do we discuss the decisions that will, in fact, transform that demand radically one way or another.
In other words, we pretend that forecasts cause decisions and outcomes, when it is decisions that cause outcomes and, in turn, condition the relevance of forecasts.
This inversion of causality is not a minor philosophical flaw. It is the reason why so many planning exercises produce plans that are already obsolete the day they are blessed, why so many S&OP cycles devolve into rituals of reconciliation instead of mechanisms of action, and why so many planners find themselves defending the sanctity of a plan that no longer makes economic sense.
Why business time is inherently asymmetric
To understand why the “neutral axis” view of time fails us, it helps to contrast film physics with business reality.
If you watch a short clip of a pendulum and you are shown both the forward and reversed versions, it is not obvious which is which. The laws do not care. In that very narrow sense, time is symmetric.
Now try the same experiment with a short clip of a warehouse. In one direction, forklifts bring pallets in, shelves fill up, orders are picked, trucks are loaded and dispatched, cash comes in. In the other direction, customers un-buy products, trucks unload returns into stores, pallets reassemble themselves, suppliers send money back.
You do not need a PhD in logistics to see which direction is real.
Economic life has an arrow of time. A decision to purchase goods today ties up capital until those goods are sold, transformed, or written off later. A pricing decision made this morning cannot be undone this afternoon for the customers who already purchased at that price. A promise of service made to a key client will cast a long shadow on what you can credibly do for others.
Every decision is a commitment that lasts for some time. During that period, some options are open and others are closed. You can divert a container already at sea only so much. You can renegotiate with a supplier only so fast. You can redesign a product or relocate a warehouse only over months or years, not overnight.
Seen from this angle, time is not a neutral axis; it is the dimension along which commitments unwind. The question is no longer, “What is the forecast for period t?” but “For how long will this decision constrain us, and how much capital will be tied up while we wait for its consequences?”
Information behaves asymmetrically as well. A piece of knowledge about your market – a competitor’s weakness, a channel that is underpriced, a pattern in customer behavior – is valuable only as long as it is not widely known and as long as conditions do not change. Once your competitors have acted on it or once the environment shifts, the same knowledge becomes a historical curiosity. Advantage has a half-life.
This is the opposite of scientific knowledge, which aims to be valid everywhere and forever. In business, the value of knowing something is tightly bound to when you know it and how quickly you can act on it. As time passes, information leaks, imitators react, and the edge disappears. No time series, however sophisticated, can capture this economic decay by itself.
Finally, the sequence of decisions matters. Today’s commitments influence which options will be on the table tomorrow. If you push a supplier too hard this quarter, you might find them “accidentally” overbooked next quarter. If you choose a rigid transport setup, you may discover later that it cannot be adjusted in time to serve a new opportunity. The set of possible futures keeps narrowing and branching as you move forward. You cannot rewind and try another path.
Taken together, these features – irreversibility of commitments, decay of informational advantages, and path dependence – make business time fundamentally asymmetric. Running the movie backward is not merely strange; it describes an impossible world.
An alternative: time as the dimension of commitments
If time is not a neutral line on which we project demand, how should we think about it?
I find it helpful to view every supply chain decision as placing a stack of coins on the table for a certain duration. Ordering a container of products is not just spending money; it is committing money and capacity for as long as it takes to manufacture, ship, receive, and sell those units. The risk is not just the size of the stake, but also how long the stake remains exposed.
A decision that immobilizes one million euros for three days is not the same as a decision that immobilizes the same million for eighteen months. Both can be profitable, but the latter leaves you far less able to respond to new information that might arrive in the meantime. A slow, heavy decision is like turning a tanker; once the rudder is set, the ship will keep swinging for a long while even if you change your mind.
In this view, agility is a measurable property: how quickly you can unwind your current commitments, free capital, and redirect it toward better opportunities. A supply chain that can rotate its inventory in a few weeks and adjust its purchase decisions daily is fundamentally more agile than one that takes nine months to clear a bet that turned out to be wrong.
Thinking in terms of commitments also clarifies the role of options. Qualifying a second supplier, even if you do not immediately shift volume to them, is a way of buying yourself the possibility of changing course later. Designing your network so that a warehouse can serve multiple regions, or a truck route can be re-optimized daily, is another way of keeping options open.
Under the forecast-and-plan doctrine, such options often look like waste. If the future is already “known” through time series, why pay to keep alternatives available? Under a commitment-focused view of time, the same options appear as insurance against the fact that the future is shaped, step by step, by your own decisions and by shocks that you cannot predict.
Forecasts still have a place here, but a humbler one. They are not blueprints of the future; they are tools to explore how different decisions might play out. Instead of asking, “What will sales be in week 37?” it is more honest to ask, “If we price higher, how might the mix of customers and baskets evolve versus if we price lower?” The forecast becomes a comparative instrument, not an oracle.
Confronting the mainstream view
Compared to this decision-centered view of time, the mainstream approach makes three decisive bets.
It bets that the future will resemble the past closely enough that time series are an adequate representation of demand. It bets that demand is largely exogenous, unaffected in any dramatic way by the company’s own future actions. And it bets that the main path to control is to commit early to detailed plans and then chase adherence.
These bets simplify life for software vendors, consultants, and bureaucracies. Time series are easy to store in databases and to display on dashboards. Plans are easy to assign to departments and to monitor via KPIs. Forecast accuracy and plan compliance are convenient levers for internal politics; they diffuse responsibility and make it hard to trace economic outcomes back to specific decisions.
But those bets are terrible when viewed through the lens of profitability and risk.
A view of time that is faithful to business reality starts from different premises. It accepts that the future is heavily shaped by the company’s own choices. It acknowledges that much of the relevant uncertainty is about structure – who buys, under which incentive, with which alternatives – rather than about fine variations around an average sales curve. It insists that the sequence and duration of commitments matter more than the illusion of a perfect plan.
From this perspective, the core questions change. Instead of asking whether the demand plan for next quarter is “locked”, we should ask how much capital we are about to freeze, for how long, and with what range of plausible outcomes. Instead of celebrating adherence to a plan conceived six months ago, we should celebrate the ability to recognize a bad bet early and to change course before the cost has fully materialized. Instead of treating pricing, promotions, and assortment as external “inputs” to the supply chain, we should bring them into the same decision cockpit, because they are the levers that shape the very demand we pretend to forecast.
None of this makes the world simpler. It does, however, make it more honest.
Closing thoughts
Time in supply chain is not just a column in a database. It is the fabric along which commitments, options, and information weave together. If we treat it as a neutral axis, mirroring the abstractions of physics, we will continue to build elaborate planning processes that regard the future as already written and then express surprise when reality refuses to cooperate.
If, instead, we accept that the future is mostly the result of what we decide today and tomorrow, the role of supply chain becomes clearer. It is the discipline of making good commitments under uncertainty, keeping our options alive where it matters, and shortening the delay between seeing and acting.
In that world, forecasts are servants, not masters. Plans are hypotheses, not commandments. And time is not something we fill with a predetermined script, but the scarce dimension in which we deploy our capital and ingenuity, one irreversible decision after another.
