Service Operating Model All articles
Service Design

The Gate Before the Service: Auditing Identity Verification as a Designed Stage

Service Operating Model

Ask a service designer to draw the customer journey and identity verification will appear as a box labelled security checks somewhere between contact and resolution. It is rarely drawn to scale. Yet for a meaningful share of customers, verification is the longest part of the interaction, the part most likely to end in abandonment, and the part where the operating model is least able to see what happened.

The reason is framing. Identity checks are typically owned by compliance and risk functions, described in policy language, and measured by fraud loss and audit findings. All of that matters. But verification is also a stage of service delivery, subject to the same design questions as any other: how long does it take, who fails it, what happens next, and is the effort proportionate to the risk being managed?

A gate, not a stage

Gates are binary. You are through or you are not. Stages are graduated: they can be timed, queued, retried, escalated, abandoned and improved.

Most UK service operations treat verification as a gate. The consequence is that verification metrics live in a risk dashboard — frauds prevented — rather than in a service dashboard covering how many legitimate customers were unable to complete their business. The two are related but not symmetrical, and the second is frequently unmeasured.

Four touchpoints where identity is tested

Map the stage before auditing it. In most multi-channel operations verification appears at four distinct points.

The fourth point deserves particular attention. Recovery is where the model meets customers at their least patient and most constrained, and where the distinction between we cannot verify you and we cannot help you collapses in the customer's mind.

Proportionate to what?

The test of proportionality has two halves, and most audits only run one.

The first is conventional: is the level of assurance matched to the risk of the action? Reading a balance and moving a large sum should not demand the same effort.

The second is usually missing: is the effort matched to the customer's capacity to complete it? A check requiring a passport, a smartphone with a working camera, a stable connection and twenty uninterrupted minutes may be proportionate to a payment change and wholly disproportionate to someone ringing to ask why a direct debit has not collected.

Where vulnerable customers leave the funnel

Verification is an exclusion mechanism by design; that is its purpose. The operating question is whether it excludes the people it is meant to exclude and no others.

Four groups recur in UK service audits. Customers with limited digital access or confidence, pushed into channels they cannot complete. Customers acting for someone else — holders of lasting powers of attorney, deputies, family members managing affairs — for whom the identity that matters is not the caller's. Customers in vulnerable circumstances, including those whose contact details have been changed by someone else, for whom standard verification questions may be unanswerable or unsafe to answer. And customers whose records are simply wrong, where a data quality problem is experienced as a security refusal.

Consumer Duty expectations around foreseeable harm, and the reasonable adjustments duty under the Equality Act 2010, both bear on this. A verification process that cannot be completed by a customer with a protected characteristic, without an alternative route that is genuinely available, is better understood as a design defect than as a customer failing.

Instrumenting the stage

You cannot audit what you do not record. The minimum viable measurement set for a verification stage is:

Running the audit

  1. Draw the stage end to end across all channels, including recovery routes, marking every point at which the customer is asked to prove who they are.
  2. For each point, record what evidence is required, what happens on failure, and who decides whether an exception is granted.
  3. Compare each requirement against the risk of the action being taken, and flag mismatches in both directions — under-checked transactions and over-checked enquiries.
  4. Sample failures and read them as evidence about the design rather than about the customer, grouping them by cause: data quality, access, capacity, acting for another, genuine fraud.
  5. Test the alternative routes. Are they published, staffed, timed and measured? An alternative that exists in policy but takes eleven working days is not an alternative.
  6. Report findings to the forum that owns the customer journey as well as the one that owns fraud, and require a single named owner for the stage.

Designing the gate

The practical conclusion is not that verification should be lighter. It is that verification should be designed, owned and measured as a stage of the service operating model, with the same discipline applied to its queue behaviour as to any other part of the journey. A gate that nobody has drawn, timed or tested is not a control. It is an unexamined exit.

All Articles

Related Articles

January's Second Peak: Auditing Reverse Logistics as a Designed Service Line

January's Second Peak: Auditing Reverse Logistics as a Designed Service Line

The Last Dropdown: How Closure Codes Decide What Your Service Model Learns

The Last Dropdown: How Closure Codes Decide What Your Service Model Learns

Neither New Nor Active: Auditing the Route Back Into Your Service

Neither New Nor Active: Auditing the Route Back Into Your Service