One Failure Away: A Ledger for the Dependencies the Board Has Not Priced
Photo: single supplier warehouse logistics dependency mapping board risk analysis UK, via d1e4pidl3fu268.cloudfront.net
Ask a board how supply chain risk is managed and the answer will usually describe a register: a matrix of categories, a colour scale, an owner against each row, a quarterly refresh. It is a reasonable instrument, and it answers a reasonable question — have we covered everything we ought to be worried about?
It does not answer the harder question: what single failure would hurt most, and how quickly would we know?
Two organisations can hold near-identical registers and carry radically different fragility. One has spread its exposure across customers, suppliers and sites; the other has thirty-eight per cent of a division's revenue resting on one plant, one system and one person who has not taken a holiday in two years. Both registers read amber. Only one of them is one failure away from a profit warning.
Categories conceal concentration
Registers are built for coverage. They are organised by type — cyber, people, supply chain, regulatory, climate — because that is how assurance is reported and how committees are chaired. Concentration cuts across those lines and is therefore routinely missed: the supplier that serves three business units, the system that underpins four processes, the individual whose departure would remove an entire capability from the organisation.
Boards tend to discover concentration in the same way: after the event, in the language of misfortune rather than exposure. The ledger is an attempt to price it beforehand.
The one-page ledger
One page, eight to twelve rows, deliberately blunt. Each row names a dependency, not a risk category, and each carries the same fields: what it would affect, whether it can be substituted, how long restoration would take, how long it would take the board to find out, what evidence exists that mitigation works, who owns it, and when it was last tested.
| Dependency | Exposure | Substitutable? | Time to restore | Detection lag | Evidence of mitigation | Owner | Last tested |
|---|---|---|---|---|---|---|---|
| Customer A (retail) | 24% of division revenue | No, in under 12 months | 9 months | 1 quarter | Account plan only | CCO | Never |
| Component B (single source) | 3 product lines | Qualified second source | 16 weeks | 3 weeks | Trial order placed 2024 | COO | Mar 2026 |
| Warehouse system | All UK fulfilment | Manual fallback | 72 hours | Same day | Failover exercised annually | CIO | Sep 2026 |
| Site, North West | 41% of output | No | 18 months | Immediate | Insurance only | COO | Never |
| Head of regulatory affairs | Licence-critical | No | 6 months | On notice | None documented | CEO | Never |
Two things become visible the moment the page is complete. First, which rows carry a time to restore that exceeds the strategic tolerance for disruption — the difference between an inconvenience and a change of plan. Second, which rows show no evidence at all: mitigation that exists in narrative but has never been exercised.
Four questions that separate mitigation from assumption
The ledger is a prompt, not an answer. Interrogate each row with the same questions.
Has the alternative been used, not merely identified? A qualified second supplier that has not received an order in eighteen months is a name on a list, not a mitigation. The question is when the alternative last carried live volume, and what it would take to do so at scale.
What is the detection lag? Many dependencies fail quietly. A key person resigns and the board learns at the exit interview; a supplier's own supplier is in difficulty and the news arrives as a late delivery. Where detection lags restore time, the board is governing blind for the interval that matters most.
Does the mitigation sit inside the same failure domain? Two suppliers on the same industrial estate, a backup data centre in the same flood plain, a deputy who reports to the incumbent — these are duplications of appearance. Genuine redundancy means the failure of one does not impair the other.
Who owns it, and can they name the trigger? Every row should have a single accountable executive and a stated condition at which the board is informed. A row owned by a committee is a row owned by nobody.
Reading the ledger for what it hides
Once the rows are populated, look for three patterns. Overlaps: dependencies that would fail together, such as a single port serving several suppliers counted separately. Tolerances breached: anything whose restoration time exceeds the period the strategy could absorb without material change. And absences: dependencies nobody thought to list because they work too well — the payment provider, the outsourced logistics contract, the one customer relationship held personally by the chief executive.
Regulated firms already do something close to this. The operational resilience regime supervised by the PRA and the FCA, with its requirement to map important business services and set impact tolerances, is a rigorous version of the same discipline, and its logic travels well beyond financial services.
Setting a boundary and a trigger
The ledger becomes a governance instrument when it is joined to a tolerance: no single dependency above a defined share of revenue, output or capability without board approval, a tested fallback and quarterly reporting. That connects directly to the risk appetite work most boards already do on paper — and gives the board something firmer to point at than a colour.
The annual surprise test
Publish the ledger, then test it. Each year, choose one dependency and run the failure as a tabletop exercise: what happens on Monday morning, who decides, how does the board learn, does the fallback function when exercised rather than described? Record the answer against the row.
A dependency the board has never tested is not a mitigated risk. It is a belief about a mitigated risk — and beliefs are what fail first.