Tolerances and Customer Harm

Customer harm is the lens through which an impact tolerance is most directly understood. A tolerance is not ultimately about systems, it is about people: what happens to customers when a service they rely on stops, degrades, or returns with wrong information. Harm can take many forms, from inability to access funds, to missed obligations, to confusion and distress caused by unclear communication, to decisions made on inaccurate information. A tolerance that does not describe those effects in terms relevant to its own customers has described a technical limit, not a harm limit, and it will struggle to guide decisions during a real incident.

The illustration here is a mortgage servicing service. When that service is disrupted, customers may be unable to check a balance, make a payment, or obtain a statement needed for another decision. Some customers may have an alternative route, while others, including customers in vulnerable circumstances, may not. The harm is therefore uneven, and the tolerance must be able to recognise that unevenness rather than averaging it away. The sections below examine the forms customer harm takes, why averaging misleads, and how customer harm should shape incident decisions.

Customer harm begins with practical loss: inability to pay, to receive, to check, or to act. A customer who cannot make a payment on time may face a knock on consequence, such as a missed obligation elsewhere, even if the original service returns soon afterwards. Harm also arises from uncertainty. A customer who cannot tell whether a payment succeeded may pay twice, or may hold back from paying at all, and either response creates a further problem. Inaccurate information adds another layer, because a customer who acts on a wrong balance or a wrong statement may make a decision that is hard to reverse.

Averaging misleads because customers are not interchangeable. A disruption that affects a broad group mildly and a small group severely cannot be judged by its average effect. The small group may include customers who depend on the service for essential needs, who have limited alternatives, or who face time sensitive obligations that others do not. If the tolerance is framed only in terms of overall impact, the severe harm to that group disappears into the average, and the firm may conclude that the limit has not been crossed while those customers are suffering intolerably.

During an incident, customer harm should shape three decisions above all: what to tell customers, what alternative to offer, and when to escalate. Communication should name the affected activity in plain terms, explain what customers should and should not do, and avoid technical language that obscures the practical effect. Alternatives should be real, not nominal: a telephone route that cannot cope with demand, or a branch route that is unavailable to the affected group, is not an alternative in any meaningful sense. Escalation should be triggered by harm, not only by elapsed time or system status, so that a disruption causing severe customer harm receives attention even if its technical footprint looks modest.

Customer harm is most often missed when the tolerance is written from the perspective of the firm rather than the customer. Firm centred language talks about system restoration, process completion, and internal queues. Customer centred language talks about what a customer can and cannot do, and what a customer understands. Both perspectives are needed, but only the customer perspective defines intolerable harm, because harm is suffered by customers, not by systems. A tolerance that never mentions a customer activity has not yet been translated into harm terms.

A reader can test a customer harm tolerance by asking what specific customer activities are disrupted, which customers are most affected, and what knock on consequences follow from delay or inaccuracy. Further questions concern communication and alternatives: what will customers be told, in what language, and what usable route will they be offered? If those questions cannot be answered from the tolerance and its supporting material, the customer dimension needs further work before the tolerance can be relied on.