Managing Supply Chain Complexity: Where It Comes From and How to Tame It

Complexity is not the same thing as size

It is tempting to equate a "big" supply chain with a "complex" one, but the two are genuinely different properties, and confusing them leads to the wrong management response. Size, in the sense of total volume — units shipped, revenue, tonnage moved — is largely a scale problem: a bigger operation needs more warehouse space, more trucks, more people, but the underlying planning task does not fundamentally change in character as volume grows, provided that volume is concentrated in a small number of products moving through a small number of routes to a small number of customer types. A business selling twice as many units of the same ten products through the same three channels is bigger, but not meaningfully more complex to plan or run.

Complexity, by contrast, comes from variety: the number of distinct products or stock-keeping units (SKUs) that must each be forecast, produced, stored and replenished individually; the number of different customer types and channels, each with their own service requirements, packaging rules or delivery windows; the number of supplier relationships and inbound routes; and the number of distribution nodes a product passes through on its way to a customer. Variety is more dangerous than volume because its effects do not add up — they multiply. Doubling the number of SKUs a warehouse carries roughly doubles the number of individual forecasts needed, doubles the number of reorder decisions, and multiplies the number of possible SKU-by-channel-by-route combinations that logistics and planning teams have to account for. A modest increase in variety along several dimensions at once — a few more products, a few more channels, a few more regional variants — can produce a combinatorial explosion in the number of distinct flows a business has to manage, long before total volume has grown by anything like the same proportion.

This distinction matters practically because the tools that work for managing size do not work for managing complexity. Adding capacity — more warehouse space, more staff, a bigger fleet — is a reasonable response to a volume problem. It does almost nothing to fix a complexity problem, because the difficulty is not that there is too much to move, it is that there are too many different things being moved in too many different ways, each requiring its own planning attention. A business that responds to rising complexity purely by adding capacity typically finds costs rising faster than the investment can explain, because the real driver — the multiplying number of combinations — has not been addressed at all.

The main sources of complexity

Complexity rarely arrives from a single source — it tends to build up gradually from several directions at once, which is part of why it is so hard to see and so hard to reverse.

Of these, product proliferation deserves particular attention because it is most directly driven by commercial incentives that point the wrong way. A sales or marketing team is usually rewarded for revenue growth and market share, and a new variant — a new pack size to win a promotion, a region-specific flavour, a private-label version for one large customer — is a fast, visible way to chase both. The team making that call rarely owns, or even sees, the downstream cost: the extra forecast line, the extra safety stock, the extra changeover time, the extra SKU-HS code combination on an import declaration. Each addition looks costless from where the decision is made; the cumulative effect, seen only much later in operations and finance, is very different.

The systematically underestimated cost of complexity

The single most important, and most consistently overlooked, fact about complexity is that its true cost is very often invisible in a business's own management accounts — not because the cost isn't real, but because of how standard, volume-based costing systems are built. A typical costing approach allocates overhead — warehousing, planning, quality and changeover time — across products roughly in proportion to volume or revenue. That method implicitly assumes every unit of every product consumes overhead at roughly the same rate, which is a reasonable simplification for a narrow, low-variety range and a badly wrong one for a wide, high-variety range.

In reality, a low-volume, high-variety SKU consumes a disproportionate amount of overhead relative to the revenue it generates. It still requires its own demand forecast — usually a less accurate one, since there is less historical data to work from. It still requires its own safety stock decision, and because low-volume items tend to have proportionally higher forecast error, that safety stock is disproportionately large relative to what the item actually sells. It still triggers a production changeover, a separate purchase order, a separate customs line, a separate warehouse slot — each taking almost the same administrative effort regardless of whether the SKU behind it sells one pallet a year or one truckload a week. When a volume-based system allocates this overhead evenly, the low-volume, complex SKU is systematically undercharged for the cost it actually causes, and the high-volume, simple SKUs are systematically overcharged, cross-subsidising the tail without anyone deciding that should happen.

The consequence is serious: a business can look profitable overall, and even show each individual product as nominally profitable, while a long tail of low-volume, high-variety lines is quietly destroying value the accounting system is not equipped to reveal. This is precisely the blind spot that total cost of ownership and activity-based costing are designed to correct — by tracing overhead cost to the activities that actually generate it (a changeover, a forecast line, a customs entry) rather than spreading it evenly by volume. An activity-based view typically reveals that a meaningful share of a company's SKU range is not covering the true cost of keeping it in the range, even though standard costing shows it as profitable.

Complexity is designed in, long before it reaches operations

One of the least intuitive but most important facts about supply chain complexity is that most of it is not created in the supply chain at all — it is created upstream, in product design, packaging and range-planning decisions, and by the time it reaches operations it is effectively locked in for the life of that product. A designer choosing to offer a product in six colours instead of two, a packaging team choosing bespoke retail packaging for a regional variant, or a product manager approving a customer-specific pack configuration to win one account — each of these decisions is made early, often quickly, by people evaluated on design appeal, brand fit or a single customer win, not on downstream operating cost. Once the product goes into production, the supply chain has to live with the resulting complexity for its entire life; a poor variant decision made in a five-minute design review can generate years of extra forecasting effort, extra safety stock and extra administrative cost that nobody who made the original decision ever sees or is accountable for.

This is why the most effective complexity-management interventions are frequently not supply chain interventions at all — they are changes to how design and range decisions are made in the first place, making the downstream cost of a new variant visible to the people deciding whether to add it. A design or range-review process that includes a genuine estimate of the incremental forecasting, inventory and administrative cost of a proposed variant — ideally drawn from the same activity-based logic described above — puts a real number next to a decision that would otherwise be evaluated purely on commercial upside, and materially changes how often "yes" is the answer.

Taming complexity without simply banning variety

Variety is not inherently bad — customers genuinely do want choice, and in many categories a broader range is a real competitive advantage, not a mistake to be eliminated. The goal of managing complexity is therefore not to minimise variety at any cost, but to capture as much of the customer-facing benefit of variety as possible while minimising the operational cost of producing and delivering it. Three levers, used together, do most of the work.

Postponement, or late customisation. Rather than building full variety into a product as early as possible, postponement delays the point at which a generic, common product becomes variant-specific for as long as is practically possible — ideally until a firm customer order exists. A product might be manufactured and held in a single generic form, with final labelling, packaging or regional configuration applied only at the last distribution point before it reaches its destination market. This directly attacks the P:D mismatch that drives the need for forecasting in the first place, described in time-based competition and lead time management: the generic, pre-postponement stock can be forecast in aggregate — far more accurate than forecasting dozens of variants separately — while only the final, low-cost customisation step has to run against a shorter-notice signal.

Platform and modular design. Where postponement delays customisation in time, platform design reduces the underlying variety that has to be customised at all, by engineering many end-customer variants to share a common base — a shared chassis, core formulation or packaging components — with only a smaller set of genuinely differentiating elements varying between them. A product range built on a shared platform behaves, for most of its pipeline, like a much smaller and simpler range than its customer-facing variant count suggests, because most of the forecasting, sourcing and production planning happens once, at the platform level, rather than separately for every finished variant.

Disciplined SKU rationalisation. Even with postponement and platform design applied well, product ranges accumulate genuinely unprofitable low-volume variants over time, and periodically pruning them is a normal, healthy part of range management. The critical requirement is that rationalisation decisions are made from true activity-based cost data — which actually charges each SKU for the forecasting error, changeover time and administrative burden it causes — rather than from gut feel, standard volume-based costing (which, as covered above, systematically hides the cost of low-volume variants), or sales-team pressure to keep every line "just in case." A review run on the wrong cost basis will tend to protect exactly the SKUs that most need to be cut, because the costing system it relies on is the one that hid the problem in the first place.

• Postponement — delay variant-specific work as late in the pipeline as possible.
• Platform/modular design — share components and base designs across as many variants as possible.
• SKU rationalisation — periodically prune genuinely unprofitable variants using true activity-based cost, not standard costing or gut feel.

These three levers are complementary rather than alternatives: platform design reduces how much genuine variety exists underneath a range, postponement delays when that residual variety has to be committed to, and rationalisation periodically removes the variants that still fail to earn their true operating cost. A business that only rationalises, without addressing how new complexity gets designed in upstream, will find the SKU count creeping back up within a year or two of every clean-up — a common, predictable failure mode of programmes that stop at the pruning step. See lean vs agile supply chains for how the choice between standardised, efficient flow and responsive, variety-tolerant flow interacts with how much complexity a given part of the chain can sensibly absorb.

The South African distributor's dilemma — and a compliance-cost multiplier unique to importing

A common and instructive example plays out inside South African distributors who carry an international principal's brand, or who supply the major retail groups. The pressure to broaden the range comes from two directions at once: the international principal often wants its full global assortment represented locally to protect brand presence, and the retail buyer, negotiating shelf space, frequently wants every colour, size and regional variant available so the category "looks complete" and no competitor can claim a gap. Both pressures are commercially rational from where they originate, and neither party bears the distributor's downstream cost of holding, forecasting and importing the resulting long tail of low-volume variants.

The distributor sits in exactly the position described earlier: sales and the principal push for maximum assortment breadth, while the real cost of that breadth — more individual forecasts, more safety stock spread thinner across more lines, more warehouse slots, more administrative overhead per rand of revenue — lands entirely on operations and finance, usually invisibly. Without a genuine activity-based view of what each additional variant actually costs to carry, the distributor has no defensible way to push back on either the principal or the retail buyer — "we think this SKU is unprofitable" carries far less weight than "this SKU costs us R[X] a year in forecasting error and safety stock we don't recover in gross margin," and only the second statement is available once the true cost of complexity has actually been measured.

South Africa adds a further, distinctly local multiplier that a purely domestic manufacturer does not face: every additional imported SKU is also an additional customs and compliance burden. Each distinct variant typically needs its own correct tariff classification (Harmonised System code), its own valuation and documentation on import, and its own line on every customs entry, with its own risk of a query or an incorrect classification. A ten-variant product line is not just ten times the forecasting and inventory cost described earlier — it is also ten separate tariff classifications to get right and ten separate places an error can occur on a single shipment's paperwork. For an SA importer, product proliferation therefore carries a compliance-cost dimension layered directly on top of the ordinary cost of complexity, which makes disciplined range management — and a genuine understanding of what each SKU actually costs to bring into the country — a materially bigger commercial issue here than in a domestic-only business.

Frequently asked questions

Is a bigger supply chain automatically a more complex one?

Not necessarily. Size is about volume — how much of the same thing is moving — while complexity is about variety, the number of distinct products, customers, channels and routes that each require their own planning attention. A business can grow substantially in volume while staying simple, if that growth comes from selling more of the same narrow range through the same channels; complexity rises when variety along several dimensions grows at once, because the number of combinations that must be managed multiplies rather than simply adding up.

Why does standard costing hide the cost of complexity?

Standard, volume-based costing systems typically allocate overhead — warehousing, planning, administration — across products roughly in proportion to volume or revenue. A low-volume, high-variety SKU actually consumes a disproportionate share of that overhead, through its own changeover, its own forecast line, its own safety stock and its own administrative handling, regardless of how little it sells. Spreading overhead evenly by volume undercharges that SKU for the cost it really causes and overcharges the high-volume lines that subsidise it, which can make a genuinely unprofitable long tail of products look profitable on paper.

Why is it hard to reduce complexity once a product is already in the range?

Because most complexity is decided early, in product design, packaging and variant decisions made by teams who are evaluated on commercial appeal rather than downstream operating cost, and who rarely see the forecasting, inventory and administrative burden their decision creates once the product is in the range. By the time operations is dealing with the consequences, the design decision is effectively locked in for the life of the product, which is why the most effective fixes often involve changing how new variants are approved in the first place, not just pruning after the fact.

Does reducing complexity mean offering customers less choice?

Not if it is done well. Postponement and platform/modular design both aim to preserve as much customer-facing variety as possible while reducing the operational cost of producing and delivering it — by delaying when a generic product becomes variant-specific, or by sharing common components across many end variants. SKU rationalisation, the third lever, does reduce the range, but is meant to target only the variants that genuinely fail to earn their true operating cost, not variety in general.

Why does complexity cost more for an SA importer specifically?

Beyond the usual forecasting, inventory and administrative cost that any additional SKU causes, an imported product variant also typically needs its own correct tariff classification, valuation and customs documentation. Every additional imported SKU is therefore also an additional compliance burden — a separate line on every customs entry with its own risk of a query or classification error — layered on top of the ordinary cost of complexity, which is not a factor a purely domestic manufacturer has to account for.

More concepts in Supply Chain Foundations & Strategy