Strategy and Network Design

Australia is in the middle of the largest rebuild of its electricity grid in living memory. Coal is leaving the system, renewables are arriving at pace, and thousands of kilometres of new high voltage transmission are being planned and built to connect the two. That task was already stretching the country's energy transmission supply chain and its skilled workforce to the limit. Then artificial intelligence arrived, and with it a wave of data centre demand that lands directly on the same wires, the same equipment order books, and the same pool of electricians and engineers.

The popular framing is that Australia is "rebuilding the grid for AI." That is not quite right, and the distinction matters. The grid rebuild is driven first by the energy transition: the retirement of ageing coal generators and the commitment to 82 per cent renewable electricity by 2030. AI and the data centres that power it are not the cause of the rebuild. They are a powerful new load arriving on top of it, compressing timelines that were already tight and intensifying constraints that were already binding.

For Australian leaders in energy, infrastructure, government, and the businesses that supply them, the practical message is blunt. The constraint is no longer ambition, policy, or even capital. The constraint is whether the physical equipment and the skilled people can be secured fast enough. This article looks at the problem through a supply chain lens: what is being built, how AI changes the demand picture, why decades of underinvestment have left Australia on the back foot, and what commercial and government organisations should actually do about it.

What is actually being rebuilt

The blueprint is the Australian Energy Market Operator's 2024 Integrated System Plan, the official roadmap for the National Electricity Market. Its central conclusion is stark: to replace retiring coal with firmed renewables, Australia needs around 10,000 kilometres of new transmission lines, with roughly 4,581 kilometres of that required just to hit the 2030 targets, according to the Department of Climate Change, Energy, the Environment and Water. AEMO costs the transmission component at about $16 billion and expects it to deliver net market benefits of around $22 billion. The urgency comes from coal: AEMO projects that 90 per cent of today's coal capacity will close by 2035, and all of it before 2040.

Funding the network side is the Commonwealth's Rewiring the Nation programme, a $20 billion pool of concessional finance and equity administered through the Clean Energy Finance Corporation. The money is now flowing into the projects that form the new backbone of the east coast grid: HumeLink and VNI West connecting New South Wales and Victoria, Project EnergyConnect linking New South Wales and South Australia, Marinus Link under Bass Strait to Tasmania, Sydney Ring South, QNI Connect into Queensland, and renewable energy zone infrastructure such as the Central-West Orana REZ.

The important context is this. Australia committed to the rebuild, set it a hard deadline, and underwrote it with public money, all before AI demand became a serious factor. The plan was always going to test the country's ability to source equipment and skilled labour. AI did not create that test. It raised the difficulty.

How AI changes the demand maths

Data centres are still a modest share of Australian electricity demand, but the trajectory is steep. In 2024-25, data centres consumed around 4 terawatt hours, about 2 per cent of the NEM, equivalent to the electricity used by more than 700,000 homes, according to analysis for AEMO by Oxford Economics Australia. The same work projects that demand triples to nearly 12 terawatt hours by 2030, lifting data centres to about 6 per cent of NEM electricity, and reaches roughly 34 terawatt hours by 2049-50, around 12 per cent of the grid.

The pipeline behind committed projects is far larger. In New South Wales alone there were 44 data centres in the development pipeline as at 31 March 2026, totalling 11.4 gigawatts, which is roughly the output of four Eraring power stations, Australia's largest coal generator. A CEFC and Baringa report released in December 2025 forecasts that data centres could account for up to 11 per cent of national electricity consumption by 2035, up from about 1 per cent today, with the sector attracting between $85 billion and $135 billion in investment and growing capacity fourfold within a decade. Around half of all planned capacity is clustered in Sydney, with Melbourne hosting roughly a quarter.

Two features make this demand hard to plan for. The first is geographic concentration. AI data centres cluster around existing fibre, land, and power, so the load piles up in a handful of corridors, principally Western Sydney and parts of Melbourne, putting acute, localised pressure on the transformers and substations that serve those pockets. The second is uncertainty. Networks are receiving connection applications well beyond what will ever be built, a phenomenon the industry calls "phantom demand," where developers lodge speculative requests to hold a place in the queue. This makes forecasting genuinely difficult and creates a real risk of over-building for loads that never materialise, or under-building for the ones that do. For anyone planning supply, the AI boom is not a single clean signal. It is a noisy, concentrated, fast-moving demand layer on top of an already demanding transition.

How Australia ended up on the back foot

Australia is not rebuilding this grid from a position of strength. It is rebuilding after decades in which it barely needed to build at all, and the supply chain muscle for large-scale grid delivery has wasted. This is the part of the story that gets least attention and matters most.

The original grid was largely built out in the post-war decades. From the 1990s, with the network broadly in place and the sector privatised and restructured, investment shifted toward maintenance and incremental upgrades rather than nation-shaping transmission. For roughly a generation, Australia simply did not do large grid construction at scale. Commentators have long observed that the country's energy infrastructure suffered from a lack of long-term investment and planning, and even the recent past tells the story: the electricity transmission sector contracted slightly through the early 2020s before the current build began to turn it around.

Two things happened to the supply chain as a result. First, the domestic manufacturing base narrowed. As recently as 2000, government analysis found that 75 to 85 per cent of Australia's transformer capacity was made domestically. A capable local base still survives, led by long-standing names such as Wilson Transformer Company, but the largest and highest-voltage units, along with many specialist high-voltage components such as HVDC cable and cable accessories, are now imported. Industry analysis has explicitly flagged that this reliance on imported high-voltage specialist items has exposed supply chain vulnerabilities. Second, the skills and project-delivery base thinned. The workforce that built the original grid aged out, apprenticeship pipelines shrank against flat demand, and the engineering and construction ecosystem that knew how to deliver large transmission contracted. The Australia Institute has documented that real productivity in the electricity sector fell by around a third between 2007 and the early 2020s, with hands-on field roles giving way to administrative ones.

This is supply chain atrophy, and it behaves the same way in every industry. When a capability is not needed for twenty or thirty years, suppliers consolidate or exit, skilled people retire without being replaced, and the institutional knowledge of how to build at scale fades. None of it vanishes overnight, but rebuilding it takes years. Australia is trying to rebuild it at exactly the moment global demand for the same equipment and people is at a record high. That is what being on the back foot means in practice, and it is why the recovery is not simply about spending money. It is about re-standing-up manufacturing, trades, and delivery capability that were allowed to wither.

The equipment squeeze: availability is now the gating factor

The grid does not run on plans and finance. It runs on transformers, high voltage cables, switchgear, circuit breakers, conductors, and steel. Every one of those items now sits in a global queue, and a country that imports the largest units is near the back of it.

The International Energy Agency's 2025 report Building the Future Transmission Grid quantified the problem. Across the global market, it now takes two to three years to procure high voltage cables and up to four years to secure large power transformers, with lead times having almost doubled since 2021. Direct current cables, preferred for the long-distance interconnectors central to Australia's plan, can take more than five years. Wood Mackenzie's market surveys through 2025 told the same story, with standard power transformers averaging around 128 weeks for delivery and the largest units stretching to four years. Prices have moved with the lead times: the IEA found power transformer prices have risen roughly 75 per cent since 2019 and cable prices have nearly doubled, driven by raw materials such as grain-oriented electrical steel, which roughly doubled in price between 2021 and 2023.

The reason is that demand is not one wave but three, all peaking at once and drawing from the same supplier base: the global energy transition, the electrification of industry and transport, and the AI data centre build-out. Global power transformer trade was worth about USD 13.5 billion in 2023, and just four countries, China, Korea, Türkiye, and Italy, supply about half of it. A buyer in Sydney or Melbourne is bidding against utilities and hyperscalers worldwide for slots in that small set of factories.

The strategic consequence is simple to state and easy to underestimate: equipment availability has replaced capital and permitting as the primary constraint on infrastructure projects. A facility that breaks ground today cannot energise on a conventional timeline, because the transformer it needs was effectively ordered years before the business case was signed. Procurement of long-lead equipment must now happen before final investment approval, not after it. The order book, not the project schedule, sets the real delivery date.

The workforce squeeze: a harder constraint than steel

If equipment is the visible constraint, the workforce is the deeper one, because a skilled tradesperson cannot be imported overnight and an apprentice takes four years to train.

Jobs and Skills Australia's report The Clean Energy Generation estimates that Australia needs around 32,000 additional electricians by 2030 to deliver the renewable target, and roughly 85,000 more by 2050, well beyond projected supply. It notes that more than half of Australia's electrical engineers were born overseas, leaving the pipeline heavily reliant on migration. Modelling of the 2024 ISP by the RACE for 2030 research centre found electricity sector employment is likely to double by 2029, an increase of about 33,000 workers in five years. The supply side is not keeping up: the Powering Skills Organisation projects an energy sector shortfall of more than 14,000 electricians by 2030, and estimates Australia needs around 20,500 apprentice electricians to commence each year through to 2030, about 40 per cent above the recent average. Close to half of electrotechnology apprentices drop out before finishing, and roughly 2.4 energy workers are approaching retirement for every new entrant under 25. This makes the shortage structural, not cyclical.

Three dynamics turn these national figures into project-level risk. The work is front-loaded into a construction peak in the late 2020s before shifting to operations and maintenance, so everyone needs the same trades in the same few years. The bulk of the work is in regional Australia, competing for the same people needed by capital-city infrastructure and, increasingly, by data centres concentrated in a few corridors. And the competition is global: the United States needs around a million additional electricians, and the IEA estimates the worldwide net zero effort requires 30 million new clean energy workers by 2030. Australia is recruiting from the same pool as everyone else, and it is no longer the automatic destination of choice. It is no surprise that an alliance of industry groups, unions, community organisations, and environmental bodies has proposed that data centres setting up in Australia be required to contribute to local energy supply and skills, rather than simply drawing on capacity others have built.

The network design challenge, and how modelling solves it

The temptation is to manage equipment and workforce one project at a time. That badly underestimates the danger, because the constraints are correlated. The same narrow window sees transmission, renewables, electrification, and AI data centres all pulling on the same transformers, cables, and crews. When a scarce resource is drawn by an entire economy at once, the risk is not just higher cost. It is that a project cannot buy the equipment or hire the people at any price within its schedule. You cannot manage correlated, systemic risk with project-by-project workarounds.

This is why the problem is, at its heart, a supply chain network design problem before it is an engineering one. The relevant network is not only the wires. It is the end-to-end equipment supply network that feeds the build: a thin set of global suppliers, oversized heavy-lift logistics through constrained ports and road corridors, staging and laydown, strategic spares, and a workforce that must be deployed across dispersed regional sites in the right sequence. Designing that network well, under genuine uncertainty, is hard for three reasons that compound each other: demand is uncertain and inflated by phantom requests, supply is long-lead and globally contested, and the work is geographically concentrated and time-compressed. A single point forecast will be wrong, and any plan built on one will fail at the first delayed transformer.

This is where scenario modelling earns its place. Rather than betting on one demand future, you model several: high and low data centre uptake, phantom versus real connections, faster or slower coal exit. You test sourcing strategies against realistic distributions of lead times rather than optimistic averages. You quantify the cost and risk of holding strategic spares against the cost of sourcing reactively when a unit fails in service. And you identify the decisions that hold up across all those scenarios, robust choices rather than ones that are optimal only for a forecast that will not eventuate.

Network optimisation then turns that insight into decisions. It tells you where to source to reduce single-point dependence, where to position critical inventory including shared transformer spares pools, how to sequence and stage builds to smooth the equipment and labour peak rather than amplify it, and how to design the inbound logistics network for equipment that needs heavy-lift handling and special road routes. Done well, this is the difference between a pipeline that overwhelms the supply chain and one the supply chain can actually serve. This is core supply chain network design, and it is exactly the kind of problem that rigorous demand modelling and supply chain analytics are built to solve, so that ordering and staging decisions are made early enough to matter rather than reconstructed after a delay has already happened.

What commercial organisations should do

For network operators, renewable developers, engineering and construction firms, equipment suppliers, and the data centre operators building their own grid connections, the practical priorities are clear. Start with an honest diagnostic of where you are actually exposed on the critical path, then act on it.

Decouple procurement from financial close. Order long-lead items, transformers and HVDC cable in particular, ahead of final investment decision, and secure factory slots and supplier relationships early. Treating procurement as a task to be sorted once funding lands is the most common way projects slip by years.

Build visibility below the first tier. The binding constraint is often a sub-component, a bushing, a tap changer, grain-oriented steel, or a cable accessory, not the headline unit. Map that exposure so a single shortage cannot quietly hold up everything else.

Treat strategic inventory as a resilience asset, not a cost. Hold or pool critical spares, especially large transformers, rather than sourcing reactively. The carrying cost is small against the cost of a stranded, near-complete project.

Plan against scenarios, not a single forecast. Use scenario modelling and network optimisation to choose sourcing, inventory, and logistics decisions that perform across a range of demand and lead-time futures, rather than optimising for one number that is almost certain to be wrong.

Build the workforce as a multi-year capability. Map the trades and engineering skills each phase needs, smooth your pipeline to avoid the worst of the boom-bust peak, plan the transition from construction crews to long-term operations and maintenance teams, and invest in apprenticeships and retention rather than assuming you can simply hire at the peak.

Diversify and qualify supply. Reduce single-source dependence, qualify alternative suppliers before you need them, and consider domestic assembly, refurbishment, and local content where it improves resilience and speed.

What government should do

Government is both the system steward and, collectively, the largest client. Its choices shape whether the supply chain can cope. The most useful interventions are supply chain decisions, not new regulation.

Coordinate and smooth the national pipeline. Sequence the build so the country does not place its entire demand for transformers, cables, and crews into the same two or three years. Smoothing the pipeline is a supply chain decision with national consequences, and it is the single highest-leverage move available.

Use collective buying power. Aggregate and coordinate procurement across projects and jurisdictions to place early, large, bundled orders that secure scarce global manufacturing capacity, through framework and advance-purchase agreements rather than each project competing alone and late.

Invest deliberately in sovereign capability. Support domestic manufacturing, assembly, refurbishment, and a national strategic-spares pool where the resilience value exceeds the lowest-first-cost case. The value of supply chain security does not show up in a standard return calculation, and judging these investments on that basis alone guarantees they never get built.

Fund the workforce pipeline at scale. Treat training capacity, apprenticeships, migration pathways, and retention as infrastructure with a four-year lead time. The crews needed at the late-decade peak have to start training now.

Improve the quality of the demand signal. Reduce phantom connection requests so planners and suppliers can invest against real demand rather than speculative queues. Better signal quality lowers the risk of both over-building and under-building.

De-risk inbound logistics. Prioritise the ports, heavy-lift corridors, and oversized-load routes that imported equipment depends on, so the last mile does not become the bottleneck after a transformer has crossed an ocean.

How Trace Consultants can help

Trace works at the intersection of supply chain, procurement, and workforce, which is precisely where the grid rebuild is constrained. We help asset owners, infrastructure developers, government bodies, and the firms that supply them turn an ambitious build programme into a deliverable one.

Supply chain strategy and network design. We model the end-to-end equipment supply network, run the demand and lead-time scenarios, and optimise sourcing, strategic inventory, and inbound logistics so the pipeline is one the supply chain can actually serve. Explore our strategy and network design capability.

Procurement and category strategy. We help organisations bring forward and de-risk the procurement of transformers, cables, switchgear, and conductors, building the supplier relationships and contracting approaches that secure capacity early rather than competing for it late. See our procurement services.

Resilience and risk management. We build the supply chain risk frameworks that make correlated equipment and labour risk visible at board level, so it can be managed deliberately across a portfolio. Learn more about resilience and risk management.

Strategic workforce planning. We map the trades and engineering skills each phase requires, model supply against a tight and ageing labour market, and design the transition from construction to long-term operations and maintenance. Explore our workforce planning capability.

Planning technology. Where the challenge calls for it, we implement advanced planning and analytics tools that bring rigour to demand scenarios, long-lead procurement, and strategic inventory, turning uncertain forecasts into early, defensible decisions. See our technology services.

For government clients and defence-adjacent infrastructure, where security of supply and sovereign capability carry additional weight, we bring sector experience through our government and defence practice, and we support major build programmes with project and change management.

Conclusion

Australia is rebuilding its electricity grid for the energy transition, and AI data centre demand has arrived on top of that rebuild at the worst possible moment for a supply chain that spent a generation winding down. The capital is largely committed and the plan is clear. What is not guaranteed is that the transformers, cables, and skilled people can be secured fast enough to deliver it, especially after decades in which the country let that capability atrophy. This is a supply chain and workforce challenge before it is anything else, and it rewards organisations that plan early and punishes those that wait.

If you are delivering grid, generation, or data centre infrastructure in Australia, now is the time to pressure-test your supply chain and workforce plans against the reality of global lead times, a thin domestic base, and a tight labour market.

Explore our supply chain strategy and network design capability →

Speak to an expert at Trace →

Related reading: Resilience and risk management · Strategic workforce planning · Trace Insights

Network Optimisation and Tool Selection: A Practitioner's View After Fifteen Years

I have spent fifteen years designing and redesigning supply chain networks, across retail, FMCG, manufacturing, infrastructure, and government, first inside large consulting and planning firms and now at Trace. In that time I have built a lot of models, sat in a lot of steering committees, and watched a lot of network studies either change a business or quietly gather dust. This is not a textbook explanation of supply chain network optimisation. It is a point of view about what separates the network work that lands from the network work that does not.

The headline of that view is simple: the optimisation engine is the least interesting part of the problem. I have seen excellent tools produce useless answers and modest tools produce decisions worth tens of millions, and the difference almost never came down to the solver. It came down to three things that get far less attention than they deserve. Whether the baseline was properly calibrated. Whether the modelling genuinely grappled with uncertainty rather than optimising to a single forecast. And whether the network decision was made together with inventory and ordering mechanics, or in isolation from them. Get those three right and the choice of tool becomes a secondary question. Get them wrong and no platform will save you. What follows is how I think about each, and where tool selection actually fits once the priorities are straight.

What supply chain network optimisation actually is, and why it so often disappoints

Supply chain network optimisation, also called network design, is the discipline of deciding the structural shape of your supply chain: how many distribution centres or plants you should have and where, which customers and regions each should serve, where inventory should sit across the network, and how product should flow from source to shelf. These decisions set the fixed footprint within which all your day-to-day planning then operates. They shape cost and service for years, which is why getting them right matters and why getting them wrong is so painful to unwind.

Network studies disappoint more often than almost any other kind of supply chain work, and I have learned to recognise the pattern early. A study kicks off with energy, a tool is licensed, a model is built at pace, and a sensible-sounding recommendation emerges, to consolidate from four sites to two, or to add a node in a growth region. The deck is polished and the savings are quantified to a suspiciously precise number. Then the recommendation goes nowhere, because when the business looks closely it does not quite trust it. The numbers do not reconcile with what finance sees. The inventory implications were waved through. The model assumed a single demand future everyone knows will not eventuate. So the boldest structural moves get deferred, and the business carries on with a footprint it has half-suspected is wrong for years.

When I unpick why these studies fail, it is almost never the mathematics. The optimisation techniques at the heart of network design are mature, and the commercial tools solve them perfectly well. The failures sit upstream and downstream of the solver: a baseline nobody calibrated, a refusal to model uncertainty honestly, a network question divorced from inventory, and an absence of the executive ownership needed to act on a structural change. The tool gets blamed, or credited, far more than it deserves. So before tools, I want to talk about the things that actually determine whether the work is worth doing.

It begins and ends with the baseline: the calibrated digital twin

If you take one thing from this article, take this. A network model is only as good as the baseline it is built on, and the most common and most expensive mistake I see is teams rushing past the baseline to get to the scenarios. Everything you build on top inherits the flaws underneath.

The baseline is what the industry now calls a digital twin: a model of your current network that reproduces, as faithfully as the data allows, how your supply chain behaves today. The major platforms lean on this language heavily, Coupa, for instance, markets a true digital twin of the extended supply chain, and the concept is sound. But a digital twin is only useful if it is calibrated, and calibration is the step that gets skimped. By calibration I mean tuning the model until it reproduces last year's reality, your actual costs, volumes, flows, and service, within an acceptable tolerance, before you trust it to tell you anything about the future. If the model cannot reproduce what already happened, it has no business forecasting what should happen next.

This is harder and less glamorous than it sounds, which is why it gets rushed. Calibrating a baseline means reconciling the model's freight cost against the freight you actually paid, lane by lane and mode by mode. It means handling and warehousing costs that reflect your real operations, not a generic rate card. It means demand that matches what you actually shipped, and lead times that reflect the variable real world rather than a single number copied from an ERP field. It means finding the awkward reconciling items: costs in the wrong cost centre, inter-site transfers nobody accounts for cleanly, peaks the annual average hides. The test I apply is blunt. Fed last year's demand, does the model reproduce last year's cost and service to within a few per cent? If not, I do not move on. An uncalibrated baseline is worse than no model, because it carries the false authority of precision. A rough spreadsheet invites scrutiny. A polished optimisation model that is quietly wrong gets believed.

The baseline also earns you the right to be believed. When I can show a CFO that our model reproduces their actuals, that it tells them what they already know to be true about last year, the scepticism drains out of the room and every number on top is easier to trust. Skip that step and the whole analysis fights an undertow of doubt. The baseline is not a chore to get through on the way to the scenarios. It is where most of the value, and almost all of the credibility, is won or lost.

Scenario modelling: the value is in the questions, not the solver

Once you have a baseline you trust, scenario modelling is where the value gets created, and the craft is almost entirely in the questions. The solver will optimise whatever you point it at. Whether the exercise is worth anything depends on whether you framed the right decisions.

The scenarios that matter are rarely just "what is the lowest-cost footprint." That is the question everyone asks first and usually the least interesting, because lowest cost in isolation is almost never what the business actually wants. The valuable scenarios explore the real decision space: at what point does consolidation start to break service for which customers; how does the optimal network change if the business grows thirty per cent, shifts toward e-commerce, or wins a major customer in a region you barely serve; what does the network need to look like if a key site is lost to a lease expiry or a disruption; how does an acquisition reshape the answer. A good scenario set brackets the genuinely consequential questions and the genuinely plausible futures, rather than running twenty minor variations to manufacture the appearance of rigour.

The discipline I hold to most firmly is to resist optimising to a single point. It is tempting to take the base-case forecast, optimise to it, and present the result as the answer. But a network tuned perfectly to one forecast is, by construction, fragile to every other forecast, and we know with certainty the single forecast is wrong. The right question is not "what is the optimal network for the expected future" but "what is the network that performs well across the range of futures we might face." That leads into the two techniques that separate serious network work from the rest: sensitivity testing and Monte Carlo analysis.

Sensitivity testing and Monte Carlo: designing for a range of futures

A network recommendation that has not been stress-tested against uncertainty is incomplete, however elegant the optimisation. Structural decisions are ones you live with for years, through conditions you cannot foresee, so the question is never just "is this optimal today" but "how robust is this to what will change."

Sensitivity testing is the simpler layer. It asks how the answer moves as you flex the inputs that matter: fuel and freight, demand volume and mix, labour and property costs, lead times. Its value is in showing which assumptions the recommendation actually hinges on. Sometimes the optimal footprint is stable across a wide range of fuel prices, so you can stop worrying about fuel and act with confidence. Other times the whole recommendation flips on a ten per cent freight movement, which is vital to know before you commit capital. I have seen recommendations that looked compelling at the base case fall apart under sensitivity testing, and I would far rather discover that in the model than two years into a property lease.

Monte Carlo analysis goes further, and for high-stakes decisions it is where the real robustness work happens. Rather than flexing one variable at a time, it runs the network across thousands of randomised scenarios, drawing each uncertain input from a probability distribution and aggregating the results to show the full range of outcomes a design produces. This reframes the decision. Instead of "design A saves two million in the base case," you get "design A saves between half a million and three million across the plausible range and never performs worse than the status quo," against "design B has a higher expected saving but a real tail risk of underperforming today if demand softens." That is a far richer basis for a structural decision, because it lets you choose a network for resilience, the one that performs well across most futures and protects you in the bad ones, rather than the one perfectly tuned to a future that will not arrive.

This matters more now than when I started, because the environment is less stable. McKinsey's research, widely cited across the industry, finds that supply chain disruptions lasting longer than a month now occur every 3.7 years on average and can cost businesses up to 45 per cent of a year's profit over a decade. When disruption on that scale is near-certain over the life of a network decision, designing to a single benign forecast is not optimism, it is negligence. A network designed with uncertainty built in is one you can defend to the board in three years when conditions have moved and the design still holds. deloitte

The decision is strategic enablement, not cost minimisation

A conviction that has only hardened with experience: the best network studies start with "what are we trying to enable," not "how do we save money." A network is the physical expression of a business strategy, and if you design it purely to minimise cost you will often design something that undermines what the business is trying to achieve.

The clearest way to see this is through the customer value proposition, because the network encodes it whether you intend it to or not. The trade-offs in network design, fewer larger sites against more smaller ones, central inventory against forward-positioned inventory, lowest cost against fastest delivery against greatest resilience, are not abstract parameters. They are decisions about what promise you make to which customers. Consolidating to two large distribution centres might be the lowest-cost answer, but if it breaks the next-day promise your best customers buy from you for, you have optimised your way out of your own value proposition. A more distributed, slightly more expensive network might be exactly right if speed and availability are how you win. There is no universal answer, because it depends on what the business is for and which customers matter most.

So I push hard, at the start of every engagement, to get the strategy and the value proposition on the table before we touch the model. Who are we serving, and what do they value: price, speed, availability, breadth, reliability? Which segments are we willing to serve differently? Where is the business heading, into which channels and regions? What is our appetite for resilience versus efficiency, knowing the two genuinely trade off and the last few years have repriced that trade-off for most boards? And where does sustainability sit, because network structure drives a large share of transport emissions, and more of the businesses I work with now treat carbon as a real constraint. Only once those are answered does optimising make sense, because they define what "optimal" means. I would rather spend a day arguing about the value proposition than a month optimising against a goal nobody has examined.

The point too many people miss: network and inventory cannot be decided apart

If the baseline is the thing teams most often skimp, this is the thing they most often get structurally wrong, and it is the most expensive class of error I encounter. Network decisions cannot be made in isolation from inventory and ordering mechanics, because they are not separate problems. They are the same problem at different time horizons, and treating them separately is how businesses arrive at footprints that look optimal on the network model and prove uneconomic in reality.

Here is why they are inseparable. Where you hold inventory is simultaneously a network decision and an inventory decision. The moment you add or remove a node, you change the inventory the network must carry, through the pooling effect: consolidating stock into fewer locations reduces total safety stock, because aggregated demand variability is proportionally smaller, while spreading stock across more forward locations increases it. This is not a second-order detail. The inventory consequence of a footprint change can be large enough to outweigh the transport and facility savings the network model was optimising for. I have seen studies recommend consolidating to fewer sites on transport and overhead savings, only for the move to collapse once someone modelled the inventory, because the forward-positioning the service promise required, combined with the suppliers' ordering constraints, wiped out the saving. The network model said yes. The inventory reality said no. Nobody had put the two in the same room.

Ordering mechanics compound this. Replenishment frequency, minimum order quantities, batch sizes, container and pallet rounding, and order policies all determine how product actually flows through the network and therefore what it truly costs to operate. A network optimised as if product flows in smooth, perfectly divisible streams will mislead you, because real product flows in lumps governed by ordering rules, and those lumps drive inventory, handling, and space. Multi-echelon inventory optimisation, which decides how much buffer to hold at each tier, is really the operational expression of a network design choice, and the two should share one model and one set of assumptions. I have written before, in our guide to supply chain planning for Australia and our work on demand, inventory, and replenishment, about how much value leaks at the seam between disciplines that should be joined. Nowhere is that seam more expensive than between network design and inventory.

The organisational version of this mistake is the one I see most. A strategy team runs the network study and hands over a footprint, while a separate planning team owns inventory on its own assumptions, and the two never reconcile. The result is either a recommendation that ignores its own inventory implications or, worse, a structural change built on transport savings that the inventory reality then undermines. My firm rule is that a network study that does not model the inventory and ordering consequences of each scenario is not finished, however polished the footprint analysis. The network question and the inventory question must be answered together, in one model, by people who are talking to each other. This single discipline, more than any tool or technique, is the difference between network work that holds up and network work that embarrasses everyone a year later.

Choosing the tool: what actually matters

Only now, with the priorities straight, do I come to tools, because this is the order in which they matter. The tool is the last and least of the decisions, and teams routinely over-invest in platform selection while under-investing in the baseline, the uncertainty modelling, and the network-inventory integration that actually determine success. The tool is not irrelevant, though, and the market has shifted enough recently to be worth understanding.

The criteria I care about, in order, are these. Can the tool support a properly calibrated baseline without taking months to stand up. Does it bring optimisation, simulation, and uncertainty analysis together, because a network you cannot stress-test is a network you cannot trust. Can it model inventory and ordering policy alongside network structure, in the same environment, so you avoid the hand-off that causes the most expensive errors. Can the people who will own it after the consultants leave actually use it. And can you re-run it as conditions change, treating network design as a living capability rather than a one-off project.

On the landscape, the picture in 2026 is genuinely in flux. For years the reference point was LLamasoft's Supply Chain Guru, which Coupa acquired in 2020 before Coupa itself was taken private by Thoma Bravo. That product is now sold as Coupa Supply Chain Design and Planning. It remains capable and proven on complex global networks, though the common view is that design investment has slowed under successive owners and the architecture is showing its age, with model construction still tied to the desktop. The most notable newer entrant is Optilogic's Cosmic Frog, built by former LLamasoft people as a cloud-native platform that combines optimisation, simulation, and risk in a single environment with a risk rating on every scenario. Alongside them sit AIMMS, with decades of pedigree in mathematical optimisation and what-if modelling, and anyLogistix, which is built on a simulation engine and is strong where dynamic, stochastic behaviour matters more than pure optimisation. Tellingly, the better modern platforms are converging on exactly the integration I have argued for: policy optimisation, which optimises reorder points and safety stock rules alongside network structure in the same model, is becoming a defining capability. The market is catching up to the idea that network and inventory belong together.

That is the case for GAINS, which is the platform we most often recommend at Trace and the one we have the deepest experience implementing. Its strength for this problem is integration. Network design sits inside a single platform that also runs demand forecasting, multi-echelon inventory optimisation, lead-time prediction, and replenishment, built around decision engineering and designed to overlay your existing systems rather than replace them. That is precisely what stops the network question and the inventory question being answered by two different tools on two sets of assumptions, which is the failure mode that does the most damage. GAINS added dedicated network design through its 2023 acquisition of 3 Tenets Optimization, and pairs it with a mature planning suite and an overlay architecture that keeps implementation cost and disruption low. For the businesses I work with, whose network decisions are inseparable from ongoing planning and inventory, that integration is the thing that matters most, and it is why GAINS is usually my recommendation. The right tool always depends on the problem in front of you, and you should choose it last, against the criteria above, not first against a slick demonstration.

Why network studies fail

To pull the threads together, here is the list I carry in my head, every item of which I have watched happen more than once.

They skip baseline calibration and build sophisticated scenarios on a foundation that cannot reproduce last year. They optimise to a single forecast and produce a network fragile to every future except the one that will not happen. They divorce the network decision from inventory and ordering mechanics, and recommend footprints the inventory reality makes uneconomic. They treat the study as a one-off rather than a living capability, so the model is obsolete within a year. They fall for tool-first thinking, investing in the platform while neglecting what actually matters. They design for cost rather than the customer value proposition, and optimise the business out of its own strategy. And they lack an executive owner and a real decision forum, so the analysis surfaces the right questions but the organisation never finds the resolve to act. Almost none of these are technical failures. They are failures of discipline, framing, and organisation, which is why I spend so little energy on the solver and so much on everything around it.

How Trace Consultants can help

At Trace Consultants, this is core ground for us, and we approach network work the way I have described it, because it is the way it actually pays off.

We build network models on baselines we have genuinely calibrated. Through our strategy and network design work, we stand up a digital twin that reproduces your actual costs, flows, and service, and we earn the right to be believed before modelling a single scenario.

We model the network and the inventory together, never apart. We bring the footprint question and the inventory and replenishment question into one model with one set of assumptions, drawing on our planning and operations capability and our multi-echelon inventory work, so the recommendation holds up against the inventory reality rather than collapsing under it.

We design for uncertainty and for your strategy. We stress-test recommendations with sensitivity and Monte Carlo analysis so the network you choose is robust across the futures you might face, and we anchor the work to your customer value proposition and growth strategy, so we optimise toward what the business is actually for.

We help you choose and implement the right tools, on the right criteria. We select for the ability to calibrate, to model uncertainty, and above all to optimise network and inventory together as a living capability. The platform we most often recommend and have the deepest experience with is GAINS, for the integration reasons above, and our view on technology sits on our technology page. Because the realities differ by sector, we bring practitioners who have done this work in your industry, whether retail or FMCG and manufacturing.

Explore our Strategy and Network Design services →

Speak to an expert at Trace →

Where to begin

If you are contemplating a network study, interrogate two things before you engage anyone or license anything. First, ask whether you can build, or have built, a baseline that genuinely reproduces your current cost and service, because if you cannot, that is the first piece of work and everything else waits behind it. Second, ask whether your network conversation and your inventory conversation are happening in the same room, with the same people and assumptions, or running on separate tracks. If they are separate, fix that before you optimise anything, because the most expensive network mistakes live in that gap.

From there the sequence is straightforward. Calibrate the baseline until you believe it. Frame the scenarios around the strategic decisions and plausible futures that matter, not a single forecast. Stress-test with sensitivity and Monte Carlo so you are choosing a resilient network rather than a fragile one. Model the inventory and ordering consequences of every option alongside the footprint. Anchor the whole thing to your customer value proposition. And choose your tool last, with a clear preference for one that keeps network and inventory together. Do that, and the recommendation will be one your board can act on with confidence.

The bottom line

If I had to compress fifteen years into a sentence, it would be this: the network is the easy part to model and the hard part to get right, and the difference is almost never the software. It is the rigour of the baseline, the honesty with which you treat uncertainty, the refusal to separate the network from the inventory and ordering mechanics that determine its real cost, and the discipline to design toward what the business is for rather than the lowest number on a slide. The tools keep improving, but they remain instruments. They amplify the quality of the thinking around them; they do not substitute for it.

The businesses that get extraordinary value from network optimisation treat it as a strategic, ongoing capability, built on foundations they trust, integrated with their planning, and owned by people senior enough to act on what it reveals. The ones that are disappointed bought a tool and hoped it would do the thinking for them. The footprint of your supply chain shapes your cost and service for years. Design it with more rigour than a single forecast and more honesty than a cost-cutting exercise, and the tool you choose to support it will be the smallest of your worries.

If you are weighing a change to your network, Trace can help you build the baseline, model it properly, and make a structural decision that holds up.

Explore our Strategy and Network Design services →

Speak to an expert at Trace →

Frequently asked questions

What is supply chain network optimisation?

Supply chain network optimisation, or network design, is the discipline of deciding the structural shape of a supply chain: how many facilities to operate and where, which customers or regions each should serve, where to hold inventory, and how product should flow from source to customer. These decisions set the fixed footprint within which all day-to-day planning operates, and they shape cost and service for years, which is why they are made using optimisation, scenario modelling, and simulation rather than judgement alone.

What is a digital twin in network design, and why does calibration matter?

A digital twin here is a model of your current network that reproduces how it behaves today in cost, flow, and service terms. Calibration is the discipline of tuning that model until it reproduces last year's real costs and service within tolerance before you trust it to evaluate future scenarios. It matters because every scenario builds on the baseline, so an uncalibrated baseline produces unreliable answers that carry the false authority of a precise model, which is more dangerous than an obviously rough estimate.

Why can't network and inventory decisions be made separately?

Because they are the same problem at different horizons. Where you hold inventory is both a network and an inventory decision, and changing the footprint changes the inventory the network must carry through the pooling effect, where consolidating stock reduces total safety stock and dispersing it increases stock. That consequence can be large enough to outweigh the transport and facility savings a network study optimises for, so a footprint chosen without modelling its inventory and ordering implications can look optimal on the model yet prove uneconomic in practice.

What is the role of Monte Carlo analysis in network design?

Monte Carlo analysis runs a network design across thousands of randomised scenarios, drawing uncertain inputs such as demand, costs, and lead times from probability distributions rather than fixed values, to reveal the full range of outcomes a design produces. It lets you choose a network for robustness, one that performs well across most plausible futures and protects you in adverse ones, rather than one tuned to a single forecast. Given how frequently major disruptions now occur, designing with uncertainty built in has become essential.

How should I choose a network design tool?

Choose it last, after you have settled the baseline, the uncertainty modelling, and the network-inventory integration, and select on criteria that matter: speed to a calibrated baseline, optimisation and simulation and risk in one place, the ability to model inventory and ordering policy alongside network structure, usability for the team who will own it, and the ability to re-run it as a living capability. The 2026 landscape includes Coupa Supply Chain Design and Planning, the cloud-native Optilogic Cosmic Frog, AIMMS, anyLogistix, and GAINS. For businesses whose network decisions are inseparable from ongoing planning and inventory, an integrated platform such as GAINS, which optimises network and inventory together, is the strongest fit.

Related reading: Supply Chain Planning: A Guide for Australia · Demand, Inventory & Replenishment: Competitive Advantage · How Advanced Planning Systems Transform Supply Chain Planning · S&OP That Actually Works in Australia

‍

How to Build a Supply Chain Business Case That Gets Approved

Most supply chain investments do not fail because the idea was wrong. They fail because the business case never cleared the room. The warehouse automation that would have paid for itself in three years, the network redesign that would have stripped millions out of freight, the planning system that would have lifted forecast accuracy: plenty of these die not at the operational level but on a finance director's desk, marked up with the words "needs more detail" or "the numbers don't stack up."

A supply chain business case is the document that translates an operational opportunity into a financial decision the board can say yes to. It is not a project plan, a vendor pitch, or a wish list. It is an argument, built on evidence, that puts a defensible number against a problem and shows how spending capital today produces a better outcome than doing nothing. Get it right and you unlock funding, momentum, and the mandate to deliver. Get it wrong and the best idea in the business sits in a drawer for another budget cycle.

This guide is for Australian operations, supply chain, procurement, and finance leaders who need to take an investment to an executive committee or board and have it approved. It covers what a strong case contains, why so many of them get sent back, and how to build one that survives scrutiny and then actually delivers the value it promised.

Why supply chain business cases get rejected

Before building a case, it helps to understand why they fail, because the failure patterns are remarkably consistent. Industry research on capital proposals suggests a large share, by some estimates around 40 percent, never secure approval, and very often the underlying project had genuine merit. The case simply did not make a compelling argument to the people holding the cheque book.

The recurring failure modes are worth naming directly. Optimistic benefit projections with no evidence behind them. Incomplete cost accounting that hides the true investment, then blows the budget mid-delivery. Missing risk analysis that pretends the path to value is smooth. Weak strategic alignment that never connects the project to what the organisation is actually trying to achieve. And the quietest killer of all: no credible explanation of how a number on a slide becomes cash in the bank.

There is also a more uncomfortable pattern, sometimes called strategic misrepresentation, where costs get understated to improve the benefit-cost ratio and slip the project under an approval threshold. It works in the short term and creates a budget crisis later. Boards that have been burnt this way become sceptical of every case that follows, which makes life harder for the next person with a genuinely good proposal.

The lesson runs through everything below. A business case is not approved because it is long, polished, or full of charts. It is approved because the decision is clear, the logic is defensible, and the path from approval to realised value is credible.

What a strong supply chain business case actually contains

A good case answers a small number of questions in an order a busy executive can follow. Strip away the formatting and the structure is always the same.

The problem, stated plainly. What is the issue, how big is it, and what does it cost the business to leave it unsolved? This is the "cost of doing nothing," and it is the single most persuasive element of most cases. If your distribution network is adding two days to lead times and bleeding service penalties, quantify that. If fragmented procurement is leaving spend unmanaged across dozens of suppliers, size it. Executives fund problems they can see and measure, not solutions in search of a justification.

The strategic link. Every dollar of capital competes with every other dollar. A supply chain case that connects to a board-level priority, whether that is growth, margin recovery, resilience, customer service, or a sustainability commitment, will always beat one framed purely as an operational tidy-up. If the organisation is chasing growth, show how the current network constrains it. If the pressure is on margin, lead with cost-to-serve.

The options, not just the answer. Both the NSW Treasury Business Case Guidelines and the Commonwealth Department of Finance investment frameworks require an options analysis for a reason: it proves you considered alternatives rather than reverse-engineering a justification for a decision you had already made. A credible case sets out a realistic longlist, narrows it to a shortlist, and includes the base case of doing nothing or doing the minimum. Boards trust a recommendation far more when they can see what it was chosen over.

The numbers, built honestly. This is the financial heart of the case. Costs and benefits over time, expressed as ranges rather than false precision, with the assumptions visible. A net present value, a payback period, and an internal rate of return where they apply. Critically, the costs must be complete: not just the capital outlay but implementation, change management, system integration, training, and ongoing operating costs. The benefits must be the kind you can actually bank, not theoretical efficiencies that never reach the P&L.

The risks and how they are managed. Generic risk registers get cases sent back. What executives want are the decision-relevant risks: delivery risk, adoption risk, the chance benefits do not materialise, cost escalation, and how each is owned and mitigated. A case that names its own weaknesses is more trusted than one that pretends there are none.

The delivery and benefits-realisation plan. The most common reason a financially sound case still gets returned is the absence of a credible path to implementation. Milestones, resourcing, dependencies, decision gates, and most importantly, how the promised benefits will be tracked and who is accountable for them after the project closes. A business case is not a one-time approval exercise. It should become the live instrument against which the investment is measured for years.

Quantifying the benefits without overselling

The benefits section is where most cases either earn credibility or lose it. The temptation is to inflate. A spreadsheet full of optimistic savings assumptions is rarely persuasive, because experienced finance leaders have seen the gap between projected and realised value too many times to take it on faith.

The discipline is to separate benefits into tiers. Hard, bankable savings come first: freight reduction from network redesign, inventory release from better planning, labour productivity from process change, contract savings from supplier rationalisation. These hit the P&L or balance sheet and can be tracked. Soft benefits come second: improved service, reduced risk, better data, greater agility. They are real but harder to bank, so they support the case rather than carry it.

The single most powerful technique is to model in scenarios rather than point estimates. Present a conservative, base, and upside case. The conservative case should still clear the hurdle rate. If your investment only works in the upside scenario, you do not have a business case, you have a hope. Showing that the numbers hold even when you are pessimistic does more to build confidence than any amount of optimism.

Tie every benefit to a mechanism. Do not claim a 15 percent inventory reduction; explain that it comes from a specific lift in forecast accuracy, applied to a specific portion of the portfolio, releasing a specific amount of working capital. The path from insight to action to outcome must be visible. When a board can see how value becomes real, they fund it.

Getting the costs right

Underestimating cost is the fastest way to destroy credibility, both at approval and during delivery. A complete cost picture covers the full life of the investment, not just the capital line.

For a supply chain investment, that typically means the capital cost itself, implementation and integration, change management and training, any transition or dual-running costs while the old and new state coexist, and the ongoing operating cost once the solution is live. A new warehouse management system is not just the licence; it is the integration with your ERP and TMS, the process redesign, the training of every user, and the support cost that recurs forever after.

Contingency belongs in the case, visibly. Leaving it out to make the numbers look better is a false economy that catches up with you the moment the first unforeseen complexity appears. A case that includes a sensible contingency and explains it is more credible, not less, because it signals that the author understands how projects really behave.

Tailoring the case to the decision and the audience

Not every investment needs the same weight of analysis, and pretending otherwise wastes everyone's time. The Australian government frameworks build this in deliberately: the level of detail required is proportionate to the size and risk of the proposal. A minor process improvement does not warrant a hundred-page case; a multi-million dollar network transformation does. Match the rigour to the scale of the decision.

Audience matters just as much as size. A CFO reads a business case differently from a COO or a board. The CFO wants defensible numbers, complete costs, and a clear view of risk to capital. The COO wants confidence the thing can actually be delivered without breaking operations. The board wants the strategic logic and the headline decision. A strong case serves all three without burying any of them, usually through a tight executive summary that states the decision and the recommendation up front, with the supporting detail behind it for those who want to interrogate it.

Lead with the recommendation. Executives assess cases quickly and they look first for the clarity of the decision, the strength of the evidence, and the credibility of delivery. Making them hunt for the ask across thirty slides is how good ideas lose momentum.

Turning the business case into a live tool

The work does not end at approval. The most valuable thing a business case can become is the standard the investment is held to over its life. The benefits projected at appraisal should be tracked through delivery and measured at closure. Too often the original projections are quietly forgotten the moment funding is secured, and nobody ever checks whether the value showed up.

This is where benefits-realisation discipline separates organisations that consistently get a return on capital from those that do not. Define the benefits precisely, assign ownership, set the cadence for measurement, and keep the case alive as a steering instrument rather than filing it away. It protects the integrity of every future case too, because a track record of delivering what you promised is the most persuasive evidence you can bring to the next ask.

How Trace Consultants can help

At Trace Consultants, we build supply chain business cases that get funded and then deliver. As a senior-led Australian advisory firm, the people who build your case are experienced practitioners who have sat on both sides of the table, not junior analysts working from a template. That matters when the case has to survive a sceptical CFO or a board that has seen optimistic numbers before.

We quantify the opportunity with your own data. We build the analysis from your ERP, WMS, TMS, and financial systems, structured to your specific cost pools and operational drivers. Whether the opportunity is in network design, cost-to-serve, inventory, or procurement, we size it with evidence rather than assumption, so the benefits in your case are ones you can actually bank. Explore our Strategy & Network Design capability for how we approach this.

We model the financials in scenarios that hold up. Conservative, base, and upside cases with the assumptions visible, complete cost accounting across the full life of the investment, and a clear view of payback and return. The kind of analysis that earns credibility in the finance review rather than losing it.

We connect the case to delivery. A business case is only as good as the value it realises, so we build the implementation and benefits-realisation logic into the case from the start. Our Planning & Operations and Procurement teams have delivered the kinds of programmes your case will need to stand behind, from forecasting and inventory through to supplier rationalisation and contract consolidation.

We bring resilience and risk into the frame. A modern supply chain case has to account for disruption and risk, not just steady-state efficiency. Our Resilience & Risk Management work helps ensure the case reflects the real operating environment rather than an idealised one.

For larger physical investments, our Warehousing & Distribution practice covers the operational design and costing that underpins a credible facility or automation case. And our wider approach to client work is built on senior delivery, solution-agnostic advice, and a standard of returning many times the value of our fees.

Explore our Strategy & Network Design capability →

Speak to an expert at Trace →

Where to begin

If you have an investment you believe in but no approved case behind it, start with the problem, not the solution. Quantify the cost of doing nothing using the data you already have. That single number, more than any vendor demo or efficiency claim, is what opens the conversation with finance.

From there, set out your realistic options including the base case, model the financials conservatively before you model them optimistically, account for every cost across the full life of the investment, and build the delivery and benefits-tracking logic in from the start rather than bolting it on. Tailor the depth to the scale of the decision, lead with your recommendation, and be honest about the risks. A case built this way does not just get approved. It gives you the mandate, the resources, and the accountability framework to deliver the value you promised, which is the only outcome that actually matters.

The difference between a good supply chain idea and a funded one is rarely the idea. It is the case behind it.

‍

‍