Commissioned by Configuration: Five CRM Settings That Quietly Write Your Service Model
Photo: Avi Rocach for Eshet Tours, CC BY-SA 4.0, via Wikimedia Commons
Every business case for a customer relationship platform says the same thing: it will give us a single view. What the case rarely says is that the same platform will also become the most consequential author of the service operating model, because its settings decide what happens to a case at each step, and those settings are changed far more often than the policy documents they contradict.
This is a governance problem before it is a technology problem. If the organisation treats configuration as an administrative function of the platform team, then the operating model is effectively being amended by people who have never been asked what the service is for.
One: routing rules decide the shape of the queue
Routing is usually set once, early, and then inherited. A case type maps to a queue; a queue maps to a team; a team works certain hours. Each of those mappings is an operating decision, and each can drift.
The classic failure is the queue with no working owner. A case is routed to a team that no longer handles that channel, and sits there until someone notices. The second is the round-robin assignment that spreads work evenly but ignores skill, so complex cases are handled by whoever is next rather than whoever is competent. The third is the routing rule that quietly overrides the documented handover, creating a second, invisible pathway alongside the one drawn on the service map.
Ask for the routing table as a document. If it cannot be produced, that is the finding.
Two: mandatory fields shape what the customer is asked
Required fields at intake are not neutral. They determine what an agent must ask, in what order, and how long the customer spends on the phone. A field made mandatory in 2022 to support a reporting requirement may still be interrogating customers today for a report nobody reads.
The quieter problem is the opposite one. Where a field is optional, it will be empty, and any analysis built on it will be built on air. Where a field is mandatory but irrelevant, agents will type a placeholder — a single letter, a dash, the word 'unknown' — and the reporting will look complete while being wrong. Sample a hundred records and count the placeholders. It is a fast, brutal test of data integrity.
There is a UK data protection dimension here worth stating plainly. Collecting more than the service needs, and retaining it because deletion requires a change request, is a familiar pattern and a poor one under UK GDPR principles. Mandatory fields should have a stated purpose, a retention rule and a review date.
Three: priority logic decides who waits
Priority is often expressed as a picklist and executed by rules that almost nobody outside the platform team has read. The result is that the queue order experienced by customers is set by a formula rather than a policy.
The more damaging variant involves service clocks. Many configurations pause the timer while a case is 'awaiting customer' or 'pending internal review'. Used well, this produces an honest picture. Used as a default, it produces a measure that flatters the operation while the customer waits — the clock stops, and so does the urgency. Under a conduct regime that asks firms to evidence fair treatment, a paused clock with no corroborating evidence of the customer's own delay is a thin defence.
Audit the pause reasons, the volumes by reason, and the average duration in each. If one reason accounts for most of the paused time, you have found a policy, not an exception.
Four: permissions decide whether the model is one service or several
Who can see a case, edit it, reassign it, close it, and export it. Permission design determines whether a handover is a real transfer of ownership or a note left in a field that the receiving team cannot access.
Where permissions are too tight, teams build shadow records — spreadsheets, side channels, informal messages — and the service acquires an undocumented second nervous system. Where they are too loose, sensitive notes about vulnerability, health or financial circumstance become visible to people with no operational reason to see them. Both are governance failures, and both are invisible in performance reporting.
Five: closure codes decide what management can know
Every case ends with a label, and the taxonomy of endings is the organisation's memory of what went wrong. A well-designed set distinguishes resolved from closed without action, withdrawn from duplicated, and unreachable from declined.
In practice, codes multiply and meanings drift. One bucket grows until it holds a third of all closures, and with it any ability to see a pattern. The test is simple: take the top three closure codes, sample twenty cases each, and read the notes. If the notes contradict the code more than occasionally, the report you have been taking to board meetings is describing a service that does not exist.
Auditing configuration as an operating control
Five artefacts make the audit tractable: the routing table, the list of mandatory fields with their stated purpose, the priority and clock-pause rules, the permission matrix by role, and the closure code dictionary. Alongside them, the change log for the past twelve months — because configuration changes are operating model changes and should be reviewed as such.
The governance ask is modest and entirely conventional: a named owner for configuration, a change control route that includes an impact assessment on service behaviour, and a standing review at the forum that already owns service performance. Configuration set in a ticket queue and configuration set in a policy document should not be able to disagree for a year at a time.
The platform is not a passive container for your operating model. It is the place where the model is executed, amended and, occasionally, contradicted. Treat its settings as decisions, and the decisions become visible.