Resilience Investment: How CFOs Value Redundancy Before the Shock

The first mistake in most resilience cases is the framing. Boards hear the word resilience and picture a rare catastrophe, so the spend is compared with a probability that feels close to zero. The evidence points the other way. Disruption is frequent enough to appear inside an ordinary planning cycle, and for a CFO the right comparison is not insurance against an improbable event but the cost of a recurring operating exposure that currently sits, unpriced, inside the profit forecast.

McKinsey Global Institute's analysis of four industries concluded that companies can now expect supply chain disruptions lasting a month or longer to occur every 3.7 years. Over a decade it estimated losses equal to almost 45 percent of one year's profits, equivalent to seven percentage points of margin on average. The study dates from 2020, but the pressure has not eased: Allianz's 2026 Risk Barometer still ranks business interruption third among global business risks, named by 29 percent of respondents.

For a finance leader in North America the implication is that resilience spend should be measured against a loss stream with a known cadence, not a hypothetical. A disruption roughly every four years falls inside a standard strategic plan and inside the tenure of most finance leaders. Allianz adds that trade restrictions have tripled to affect an estimated US$2.7 trillion of merchandise, nearly 20 percent of global imports, so policy now generates disruption as readily as weather or fire.

Start by restating the exposure in the language the board already uses. Take the last three disruptions your business actually experienced, however minor, and compute the margin lost, the expediting cost and the customer attrition. Then present the McKinsey cadence as an external benchmark. The aim is a single page that replaces a rare event with a recurring exposure the company currently self-insures at an unknown price. That reframing changes how every later number is read.

A caution on use of external benchmarks: they establish plausibility, not precision. The McKinsey figures come from expert estimates in four industries and a particular period, and a pharmaceuticals business will differ from a metals processor. Say so on the page. Boards are more persuaded by a CFO who states the limits of a benchmark and then shows the company's own loss history beside it than by one who presents an industry number as though it were a forecast for the firm.

Finance models love averages, and averages are precisely what make redundancy look indefensible. If a disruption costs a modest amount per year once spread across a decade, then any standing buffer that costs more than that amount appears to destroy value. The flaw is that losses do not arrive evenly. They arrive in concentrated episodes in which margin collapses within weeks, and the consequences of that concentration, not the annual average, are what threaten the business.

McKinsey Global Institute makes the point directly. It found that a single prolonged production-only shock would wipe out between 30 and 50 percent of one year's EBITDA in most industries, while the ten-year average loss was almost 45 percent of one year's profits. Both figures describe the same exposure; only the first tells a CFO what the balance sheet must survive. The averaged number is a budgeting convenience, and the episodic number is the risk.

The implication is that the true cost of disruption includes second-order effects that averages omit. A shock of that scale in a single year can strain leverage covenants, delay dividends, force working capital drawdowns and push management into unfavourable supplier or customer concessions. None of these appears in an expected-loss calculation, yet each is a cost the shareholders bear. Redundancy is cheaper than it looks once these consequences are priced in.

Replace the single expected-loss line with a distribution. Build a base case, a severe case and an extreme case, each with a duration, a lost contribution margin and a leverage impact, and show them beside the annual premium of the proposed buffer. Ask the board a different question: not whether the buffer pays back in an average year, but what the company can afford to lose in a bad one.

There is a governance point as well. Averages tend to survive because nobody owns the tail. Procurement owns cost, operations owns delivery and treasury owns liquidity, so no one is accountable for the interaction between them during a long outage. Naming the executive who owns the severe scenario, and asking them to present it, is often the quickest way to move the discussion from abstract concern to funded action.

A buffer is best understood as an option. The premium is the carrying cost: inventory held, a second supplier qualified and kept warm, spare capacity paid for and unused. The payoff is the margin preserved when the primary route fails. The strike is the trigger that tells operations to draw on the buffer. Seen this way, a resilience investment stops being an unexplained cost line and becomes a purchase with identifiable terms that can be compared.

This is not only our framing. Researchers at the MIT Center for Transportation & Logistics argue that resilience should be valued on the value of maintaining operations when adversity strikes, rather than on the cost of a supply chain disruption, and propose applying real options, a method drawn from banking. The distinction matters. Valuing avoided cost invites arguments about probability; valuing continued operations asks what revenue is worth protecting and for how long.

For the CFO the options lens clarifies three decisions. Which nodes deserve protection, because the option is worth most where recovery is slow and margin is high. How long the protection must last, because the tenor determines the premium. And who may exercise it, because an option nobody is authorised to use is a cost with no payoff. Dual sourcing with a minimum volume commitment, reserved capacity and safety stock are all options with different decay profiles.

For every material node in the network, record three numbers: the annual premium, the days of revenue the buffer protects, and the lag between a trigger and the buffer becoming usable. Rank nodes by protected days per unit of premium. The ranking will show that a few cheap actions deliver most of the protection and that some expensive ones deliver little. That ranking is the substance of the capital case.

Options also have a useful property in negotiation. A supplier that knows you hold a qualified alternative, even one you rarely use, tends to behave differently from one that knows you are captive. Part of the premium is therefore recovered through better terms on the primary contract. Finance should ask procurement to estimate that effect separately, and to be cautious about counting it, because it is real but difficult to verify and easy to overstate.

The size of any resilience investment should be bounded by what it protects. The relevant quantity is the contribution margin lost while a node is down, multiplied by the time it takes to restore supply, less whatever the buffer allows the business to keep earning. This is a scenario calculation rather than a statistical one, and it can be done node by node with data finance already holds: margins by product, lead times by part and recovery estimates from operations.

Anchor the scenarios in the external evidence. McKinsey Global Institute defined material disruptions as those lasting a month or longer and found them arriving every 3.7 years on average, with a prolonged production shock erasing between 30 and 50 percent of a year's EBITDA. These figures are not forecasts for your business, but they are a defensible starting range for duration and severity, and they stop the debate from beginning at an arbitrary number.

The decision rule that follows is simple and useful. Over one expected disruption cycle, the cumulative premium paid for a buffer should be lower than the margin it is expected to protect. Where recovery time is long and the buffer is cheap, the test is passed easily. Where recovery is quick or margin thin, it fails, and the company should accept the risk deliberately. Both outcomes are legitimate; the failure is leaving the choice unmade.

Give FP&A ownership of the scenario model and refresh it every quarter alongside the forecast, so that resilience is reviewed on the same cadence as everything else. Pair each scenario with a named operational owner who validates recovery times. The first iteration will be rough. What matters is that the board sees a consistent method applied to every node and can challenge the inputs rather than the conclusion.

Be wary of false precision in the probabilities. The cadence quoted by McKinsey is an average across industries, and the true frequency for a given node may be higher or lower. A sensible approach is to run each scenario at several frequencies and report the break-even: how often must the disruption occur for the buffer to pay for itself? If the answer is far less often than the benchmark suggests, the case is robust, and the board can see why.

Before buying physical buffers, buy information. A company that cannot see where a disruption is forming cannot exercise any option, and will pay for protection it does not need while missing exposure it cannot see. Visibility is also unusually cheap compared with inventory or duplicate capacity, which makes it the first line of the resilience portfolio and the one with the clearest return.

McKinsey's 2025 supply chain risk survey of 100 companies found that 95 percent now have visibility into at least tier-one supplier risks, but that visibility extends into tier two or beyond for only 42 percent. The same survey found the share of companies planning major investments in digital supply chain systems fell from 47 percent to 25 percent in a year. Meanwhile, of those facing tariff impacts, 45 percent are increasing inventories, 39 percent are dual sourcing and 33 percent are developing nearshoring or onshoring plans.

The pattern suggests companies are reaching for the most tangible buffers first and the least tangible information last. Inventory is easy to explain and easy to audit, but it ties up capital and protects only against shocks the company has anticipated. Where the real concentration risk sits two tiers down, a warehouse of finished parts is an expensive way to be surprised. The CFO is the right person to challenge that sequencing.

Fund tier-two mapping for the highest-spend and highest-margin nodes first, and treat it as a resilience line with its own return metric: the share of revenue-critical components whose sub-tier sources are known. Protect the digital budget from the next cost-cutting round, since it determines how much physical buffer the company needs. Report progress to the audit or risk committee each quarter.

Be honest about what visibility cannot do. Mapping sub-tier suppliers tells you where concentration sits; it does not remove it. Some mapped exposures will turn out to be unavoidable, such as a single global source for a specialist material, and the right response is then an inventory or substitution plan rather than more analysis. The map earns its cost by directing physical buffers to where they matter, and by showing which proposed buffers are protecting the wrong place.

In North America the most visible source of disruption is now trade policy, and its effect on margin is immediate rather than speculative. A resilience case that ignores tariffs will look detached from what the board already discusses, while one that treats tariff volatility as a scenario input will connect to existing concerns. The practical task is to show how the same investments protect against both physical and policy shocks.

McKinsey Global Institute reports that the average effective US tariff rate jumped from 2.4 percent in late 2024 to about 22 percent in early April 2025, before trade agreements and policy adjustments reduced it to about 15 percent by year-end. McKinsey's supply chain survey found that 82 percent of respondents said their supply chains are affected by new tariffs. In its own April 2026 survey, McKinsey found that 37 percent of CFOs cite geopolitical instability and 32 percent cite trade policy as the biggest risks to growth.

Pricing alone will not absorb that volatility. Gartner polled 192 CFOs and found the average expected pass-through of tariff increases to customers was about 73 percent, leaving the remainder on margin, while 48 percent were already reengineering the supply chain through alternative components and raw materials. The signal is that finance leaders expect to pay part of the cost themselves and are looking for structural responses. Optionality, not a single optimised route, is the sensible structural answer.

Test every proposed resilience investment against at least three policy states: the peak rate, the current rate and a reversal. An investment that pays off only in one state is a bet, not an option. Prefer actions that retain value across states, such as qualified alternative sources, flexible final assembly and contract terms that share tariff risk. Present the results as a range, and be explicit about which assumptions drive it.

One further complication is timing. Tariff changes can be announced with little warning and reversed just as quickly, which punishes investments that take years to complete and rewards those that can be switched on or off. Finance should therefore ask for the lead time of each proposed action alongside its cost. A cheaper measure that takes two years may be worth less than a dearer one that can be activated in a quarter, because the policy state it was designed for may no longer apply.

Resilience programmes fail when they are funded as projects and then forgotten. Buffers decay, supplier relationships lapse and the risks they were built for change shape. A portfolio view treats each protection as a holding with a carrying cost, a review date and a trigger for release or disposal. It also forces the trade-off between resilience and efficiency into a single decision forum rather than leaving it to compete across functions.

The starting point is not strong. Allianz's 2026 Risk Barometer found that only 3 percent of respondents rate their supply chains very resilient, with 32 percent calling them resilient and 58 percent somewhat resilient. At the same time, McKinsey's April 2026 survey found that nearly two-thirds of CFOs are responding to geopolitical uncertainty by increasing cash and liquidity buffers. Cash is the default hedge, because it is within the CFO's gift and easy to explain.

Liquidity buffers and operational options address different problems. Cash helps the company survive a shock, but it does not keep customers served or protect market share while production is down. A finance function that holds only the first has bought survival, not continuity. The board should see both on the same page, with their respective premiums and the days of revenue or weeks of runway each purchases.

Set an annual resilience premium budget and release it in stages against explicit tripwires: a supplier's credit deterioration, a port or lane closure, a policy announcement. Report days of revenue protected each quarter, and retire protections that no longer earn their carry. Discuss contingent business interruption cover with insurers as a complement, not a substitute, so the programme is designed as one system rather than three.

Finally, protect the programme from its own success. If the buffers work and nothing visibly goes wrong, the premium will look like waste at the next budget round. Record near misses and the revenue they would have cost, and report them with the same regularity as the premium. A resilience programme that cannot show what it prevented is always the first to be cut, and cutting it restores the original exposure at the moment the company has forgotten it.

Our Strategy & Transformation work helps finance teams build the scenario model, the option-pricing logic and the board paper that turns resilience from a cost argument into a capital allocation decision. We work with the CFO and operations leads to validate the inputs, and we support the process rather than supplying a ready-made answer.