A bank can answer calls quickly and still leave a bereaved customer unable to access essential funds. It can send every required document while a customer does not understand an overdraft charge. Banking compliance therefore needs evidence about what happens to people, alongside evidence that procedures were followed.
For UK retail business within scope, the FCA's PRIN 2A.9 requires monitoring of consumer outcomes. The FCA's vulnerability findings emphasise defining good outcomes, investigating differences and assessing improvements. The framework below is an implementation proposal for turning those expectations into a usable testing programme.
Start with a journey and an outcome
Choose a specific journey: reporting a bereavement, requesting accessible communications, disputing a payment or seeking financial difficulty support. Define the intended outcome before selecting metrics. For bereavement, this could be accurate information, proportionate evidence requests, access through an appropriate channel and completion without avoidable distress or repeated explanations.
Then identify the points where a customer could be harmed. An online-only upload step may exclude someone who cannot use it. A rigid script may request information already supplied. A case can be administratively closed while the underlying customer need remains unresolved. These are testable failure modes, not simply training topics.
Compare like with like
An aggregate satisfaction score can conceal unequal outcomes. Compare customers using the same product and journey, considering support need, channel, complexity and stage. Avoid interpreting every numerical difference as proof of unfair treatment: case mix and small samples can distort comparisons. Equally, do not dismiss an adverse difference merely because its cause is not yet established.
A useful dashboard pairs outcome and process measures. For example, report unresolved needs after first contact alongside repeat-contact rates; delays to access alongside document rejection; and complaint themes alongside the time taken to correct the problem. Keep denominators visible. A rate based on ten cases should not be presented with the same confidence as one based on ten thousand.
Use sensitive information deliberately
Support should not depend on unnecessary disclosure. Record what staff need to deliver an adjustment, who can access the information and how it will be kept current. The joint FCA–ICO statement explains how outcome monitoring and data protection interact; it does not require firms to systematically collect new protected-characteristic data to meet monitoring requirements.
My recommended design is to separate an operational support instruction from detailed personal history wherever possible. A communication preference may be enough for a colleague to act. Compliance, privacy and customer-support specialists should agree the information needed for each purpose, rather than letting free-text case notes become a permanent repository of intimate details.
Test the whole case
In a hypothetical review, a bank samples 60 bereavement journeys across branch, telephone and digital channels. Eight customers repeatedly supply the same document. Averages look acceptable because most cases finish quickly, but the affected subgroup experiences a long delay.
The review should trace those eight cases through intake, indexing, transfer and closure. Suppose the cause is an attachment field that does not move between teams. More empathy training alone will not fix it. The bank should correct the handoff, identify other affected customers and test whether the repair prevents repeat requests. The example illustrates diagnostic reasoning, not a regulatory sample-size requirement.
Close the loop with evidence
Assign each improvement an owner, delivery date and outcome-based success criterion. Re-test the affected journey after enough cases have passed through it. Record adverse side effects: a simplified process that speeds completion but loses identity checks is not a complete improvement.
Give the governing body a short decision record: the harm identified, the people affected, the cause, the remedy and the evidence still missing. Include cases that remain unresolved. This gives senior management something more useful than a green status label: a clear basis for deciding whether customer needs are now being met and whether further action is necessary.

