CYBER RESILIENCE

When your supplier’s supplier fails

Sharing a supplier with a prominent target can change a firm’s cyber exposure. Find out what you share, who gets help first and whether your fallback works.

OpinionCyber resilience8 min read

THE EXECUTIVE VIEW

Three takeaways

  1. A prominent customer may make a supplier more attractive to attackers. Shared systems, access and recovery capacity determine what another customer stands to lose.

  2. Differently branded suppliers can rely on the same identity, hosting or support provider. Test whether they can actually fail independently.

  3. Agree how much urgent work you can process safely while a supplier is unavailable, then rehearse the pause, recovery and reconciliation.

When the power failed at a Google Cloud availability zone in London on 20 October 2026, some UK financial firms felt the effects through suppliers that used the cloud service themselves. Their own relationship with Google was only part of the picture. Banking and payments services were affected, and some firms needed more time to restore their operations after reconnecting. A supplier’s service can be back while customers are still unable to complete the work they came to do.

The UK financial authorities’ 4 November letter on the Google Cloud outage (opens in a new tab) describes a power failure. A cyber attack could expose the same dependencies and leave firms with a harder decision: is it safe to trust the data, access or transactions that are coming back? That makes the question especially important for a smaller firm using the same providers as much larger institutions. Could someone else’s prominence become your problem?

Who else uses our supplier

Yes, it is worth asking. A prominent customer can make a particular route into an organisation more attractive. But sharing a supplier with a global bank does not establish that a fund administrator, fiduciary firm or payments business is more likely to be breached. The useful question is whether the attacker can reach other customers through the systems, software or people that support the target. Two firms buying different products from the same company might share very little. Two firms buying from different companies might depend on the same platform.

The US Government Accountability Office’s SolarWinds testimony (opens in a new tab) illustrates the difference between exposure and compromise. The compromised Orion software reached an estimated 18,000 customers. That figure describes distribution; it cannot tell a board which customers experienced further intrusion. The lesson is to ask separately what can reach you through the supplier and what evidence you would need to determine how far it went.

The initial motive also matters less to recovery than it does to an investigation. A supplier can be deliberately targeted as a route to one customer, or it can be compromised because criminals can reuse the same access across many businesses. Either way, the provider may need to suspend a shared feature, revoke access or restore the system for several customers at once. A smaller firm can lose the ability to serve clients even when its own data was untouched.

Cybercrime can scale and so can its effects

On 12 November 2026, the US Department of Justice announced an international operation against Gootloader infrastructure (opens in a new tab). It attributed at least 25,000 computer infections worldwide to the malware since 2018 and linked it to ransomware attacks against at least 44 US victims. Those are different measures. The case shows how infrastructure used to gain access can support ransomware activity across many victims over time.

There is a parallel on the defensive side. In Verizon’s 2026 Data Breach Investigations Report (opens in a new tab), 48% of breaches in the dataset involved a third party, up from 30% in the previous edition. That comparison describes the incidents contributed to the study. It is useful evidence of how often third parties figure in observed breaches; it does not estimate the odds that any particular supplier will be attacked or prove a rise of the same size across the economy.

A supplier can spread both the benefit of good engineering and the consequences of a bad day. A large provider may give a mid-sized firm stronger monitoring and recovery capability than it could build for itself. The trade-off is that more firms may share the components the provider depends on. And recovery can queue: a contractual service target says little about how many customers can get specialist support when all of them need it at once.

The 17 November Optimizely incident report (opens in a new tab) shows why product boundaries matter. Optimizely reported that unauthorised code had been added to some customers’ Web Experimentation projects the previous day. Exposure depended on what an affected page did and how it was configured; the supplier separated this product from others it said were unaffected. For a financial firm, that makes the inventory of what actually runs on a payment page more useful in the first hours than a register of supplier names.

Ask about what is shared

Start with a service the board cannot afford to lose. Ask the supplier to describe the customer mix for the product and environment you actually use. An anonymised breakdown by sector and deployment type may be enough to start the discussion. A list of famous customers will tell you little without the details of how the service is delivered.

Find out which components are common across customers: identity and administration, software releases, hosting, support access and backup restoration. Ask how a problem affecting one customer is contained and which actions would affect everyone. Request evidence that fits the question, such as a summary of a joint recovery exercise or an assurance report whose scope covers the product and the shared components. Record the dependencies the supplier cannot disclose or substantiate. Those gaps matter most where you have no independent means of continuing.

Then ask what happens when several customers want the same help. How quickly will your firm learn that its service might be affected? Who can provide a view of your own exposure? How does the supplier decide which services to restore first, and how would it support a smaller customer during a prolonged incident? Ask which parts of that answer depend on a subcontractor. A relationship manager’s promise carries more weight when a timed exercise or a contractual escalation route backs it up.

The Financial Stability Board’s third-party risk toolkit (opens in a new tab) gives financial firms a reference for identifying critical services and managing dependencies. Start where client harm or financial loss would occur first. A smaller firm can get further by testing a handful of services with their owners than by sending the same questionnaire to every supplier.

Make an independent fallback earn its name

Take a fund administrator whose client portal and outsourced transfer agent both use the same upstream identity service. The contracts sit in different files and the vendors have separate continuity plans. If that identity service is suspended during a cyber incident, neither route may be usable. If staff then accept payment instructions through an unfamiliar channel, the workaround can turn a service interruption into a fraud loss. The critical dependency is the ability to authenticate an instruction and record it once, from receipt to settlement.

An exercise should take both routes away on a dealing day and run long enough for a backlog to build. The team should try to reach authorised contacts through numbers verified before the incident, record any permitted instructions in a controlled register and retain dual approval. Measure how many requests can be handled safely, who gets priority and what must wait. Then bring the systems back and make the team reconcile queued and completed instructions against trusted records before it releases the backlog. A phone tree and a spare spreadsheet are only useful if people can make that process work at realistic volumes.

That exercise should also expose the economics. A workaround that handles ten instructions an hour may be acceptable for an hour and untenable by the end of the day. Dealing cut-offs, settlement obligations and the ability to reach signatories can make harm rise faster than the queue itself. Agree a threshold at which the firm tells clients what has stopped and gives them a reliable time for the next update. If the administrator cannot verify the integrity of its records when service resumes, the clock for safe recovery is still running.

Where an independent route is affordable, prove that it works without the shared dependency. Where a fallback cannot manage the volume, agree a controlled pause: who can call it, how customers will be told what to expect and which integrity checks must pass before restart. A new contract is justified when it buys measurable continuity or a credible exit. It adds little if the replacement relies on the same upstream service or if the firm cannot move its data and work in time.

Give the board a decision

Within 30 days, the COO and the service owner should map one important service from the customer outcome back through the suppliers and subcontractors it needs to operate. Within 60 days, Procurement and Resilience should establish where the primary and fallback routes share technology or recovery capacity and identify what they still cannot verify. Within 90 days, the service owner should run the failure exercise and bring the board the results: safe throughput, time to restart, unresolved transactions and the customer exposure that remains.

The board can then decide whether to fund an independent route, agree a different recovery commitment with the supplier or accept a time-limited interruption on explicit terms. The best question to end on is concrete: if our supplier is recovering several customers at once and we are still waiting for help, what can we safely do for ours?