The contact center was solving problems the business created.

At a glance

Organization

Large service organization with a high-volume assisted-service operation.

Situation

1.2M assisted customer contacts annually at $9.25 fully loaded per contact — $11.1M of annual assisted-contact cost, of which $6.45 per contact is avoidable at volume.

Complication

Every improvement lever in play — handle time, productivity, self-service, automation, staffing — accepted the volume of demand as given.

Entry point

Zero-Based Transformation™, with GET, SORT and DO the most visible stages.

Result

36% of demand classified as failure, repeat or potentially suppressible; 169,000 contacts modeled as removed at source; $1.09M of modeled annualized run-rate cost reduction against a $1.57M fully loaded equivalent that was not claimed.

The levers were the usual ones. The demand was not.

Management was working the standard program: reduce handle time, raise agent productivity, push volume to self-service, automate the repeatable, and match staffing more tightly to arrival patterns. All five are legitimate, and the operation was competent at all five.

What they share is an assumption. Each takes the arriving volume as a given and works on the cost of serving it. Handle time reduces the unit cost of a contact. Self-service moves a contact to a cheaper channel. Automation removes the human from part of it. None of them asks whether the contact should have existed.

That assumption is usually invisible, because the contact center is measured on how it handles demand and has no authority over the parts of the business that create it.

A material share of demand was generated internally

Read by origin rather than by handling metric, the volume separated into categories that behave very differently. The distinction that matters commercially is between demand that is potentially suppressible and demand that is genuinely necessary — and then, within the first group, between what can and cannot responsibly be removed:

• Necessary demand — the customer legitimately needs a person, and the contact is the service working as intended.
• Preventable demand — something upstream failed, and the contact exists to repair it.
• Repeat demand — the first contact did not resolve the issue, so the customer came back.
• Process-generated demand — a step in a process reliably produces a question or a confusion.
• Policy-generated demand — a policy is written in a way that requires the customer to call.
• Unresolved or exception demand — the case falls outside a defined path and has nowhere else to go.

Four of those six are consequences of how the enterprise operates. They arrive at the contact center, are measured there, and are managed there — but they are not produced there.

Why does this contact need to exist at all?

Once demand is read this way, the improvement question changes shape. A repeat contact is not a productivity problem; it is evidence that a resolution path failed. A policy-generated contact is not a staffing problem; it is a policy decision with an operating cost attached that nobody priced when the policy was written.

Why efficiency programs stall here

A contact center can absorb a great deal of efficiency improvement and still see cost rise, because the two forces are independent. Unit cost falls while volume grows, and the volume growth originates outside the operation’s control.

Deflection has the same limitation in a subtler form. Moving a preventable contact to self-service does not remove the failure that generated it; it relocates the customer’s experience of that failure to a cheaper channel, where it often produces a second contact anyway.

This is why service organizations can run world-class handling metrics and still be expensive. The operation is optimizing the response to a demand signal that the rest of the business keeps generating.

Read the demand before optimizing the channel

The work was to establish where demand originated, separate what could responsibly be removed from what could not, and then change the upstream conditions rather than the handling.

First, classify demand by origin

A system-level diagnostic classified contacts into the six categories above. The classification is deliberately about cause rather than topic — two contacts about the same product can have entirely different origins, and only the origin tells you what to change.

Second, separate avoidable from responsibly removable

This is the step that determines whether the number is defensible. Of the demand classified as failure, repeat or potentially suppressible, a substantial share cannot responsibly be suppressed: regulatory obligations, deliberate policy positions, genuine case complexity, and situations where a human conversation is the right service even if a system could technically handle it. That share was excluded from the addressable pool before any target was set.

Third, change the cause

For the remaining pool, the intervention sat upstream: process steps that reliably generated confusion, policies written without their contact cost priced in, and resolution paths that did not close. The contact center was the place the problem was measured, not the place it was fixed.

Fourth, hold a realistic modeled removal rate

Not all addressable demand can be removed in one cycle. A 70% removal rate against the addressable pool is the modeled deliverable, rather than the 100% that a business case would prefer.

What could responsibly be removed

How the cost actually left the base

A contact that disappears does not take all of its allocated cost with it. Of the $9.25 fully loaded cost per contact, $4.10 is variable — handling time, communications and transaction cost — and leaves the base with the contact. A further $2.35 is semi-variable: supervision, quality assurance and overflow vendor capacity, which reduces only if volume reduction is converted into an actual staffing decision. The remaining $2.80 is fixed allocation across technology, facilities and management, and does not move at this scale of reduction.

The removal converted as follows: 169,344 contacts at an average handling and wrap time of approximately 7.5 minutes releases roughly 21,200 hours, or about 12.4 full-time equivalents of capacity. Nine positions are modeled as removed through attrition without backfill and overflow vendor hours reduced, which is what realizes the semi-variable element in the model. The residual 3.4 FTE of released capacity is modeled as absorbed into service-level improvement rather than taken out, and is therefore reported as capacity, not as saving.

Basis of the percentages: contact classification is applied to a twelve-month population of assisted contacts, coded by origin rather than by topic; the 36% classification and the 44% non-suppressible partition are modeled, and the 70% removal against the suppressible pool is a modeled outcome rather than a delivered result. Cost elasticity per contact is a modeled split of the fully loaded rate. No saving was claimed against the full 432,000, and the $1.57M fully loaded equivalent is shown for reference only — the claimed economic result is $1.09M.

The reduction came from upstream

The modeled intervention makes process and policy changes at the points where contacts are generated, owned by the functions that own those processes rather than by the service operation.

Contact reduction stopped being a contact-center target. It became an operating-model outcome with named owners outside the channel — which is the only place the change could actually be made.

The addressable pool was tracked separately from total volume, so that seasonal or growth-driven increases in necessary demand could not be mistaken for program failure, and vice versa.

The results

169,344 contacts are modeled as removed annually, publicly rounded to 169,000. That is the modeled operating outcome. The modeled economic result is $1.09M of annualized run-rate cost reduction — the avoidable $6.45 per contact, realized through nine positions not backfilled and reduced vendor overflow hours. The fully loaded equivalent of $1.57M was not claimed, because $2.80 per contact of fixed allocation did not leave the cost base.

Claimed against the full 432,000 contacts at the fully loaded rate, the same program could have reported roughly $4M. That number would not have survived the first serious question about regulatory demand or about which costs actually left the base — and the credibility lost would have cost more than the difference.

Why the result held

Because the cause was removed rather than the contact deflected, the volume did not return through another channel. A contact that is never generated cannot reappear in email, chat or a second call.

The non-suppressible share was excluded before targets were set rather than negotiated afterwards, so the program was never in the position of defending a number it had already missed.

What this means if you run a service operation

If handling metrics are strong and cost is still rising, the operation is not the problem. The demand signal is, and it is being generated somewhere with no visibility of what it costs.

Evidence basis

Benchmark research supports both the scale of failure demand and its cost consequence.

• SQM Group, First Call Resolution: a comprehensive guide (benchmarking of 500+ North American contact centers). For calls that are not resolved on first contact, the agent is the source of error in roughly 38–40% of cases — meaning the majority originate elsewhere in the organization. SQM links each 1% improvement in FCR to a 1% improvement in customer satisfaction and to lower operating cost.
• SQM Group, What is a good First Call Resolution rate?. Places a good FCR rate at 70–79% and world-class at 80% or higher, and reports that only about 5% of benchmarked contact centers reach the world-class standard.