Want help applying this guide?Ask the DripTell team
Customer service cost per resolution is the total cost of support work in a defined period divided by the number of unique customer issues that were fully resolved and stayed resolved through a fixed check window. The arithmetic is simple. The difficult part is making both sides of the division honest.
If the cost pool contains only agent salaries, the result is too low. If the resolution count includes duplicates, abandoned conversations, premature closures or cases that quickly reopen, the result is also too low. A useful measure includes the whole operating cost and counts verified outcomes rather than closed records.
Start With the Outcome You Are Paying For
Microsoft's workplace service pattern moves measurement away from tickets handled and toward resolution, satisfaction and cost per resolution. Its customer service value blueprint adds the baselines needed for a defensible cost model, including contact volume by channel and intent and fully loaded representative cost per hour. It also keeps resolution, abandonment, first contact resolution and customer satisfaction visible.
- 1Set the periodUse one complete cost and resolution period.
- 2Build the cost poolInclude direct work, enabling work, technology, overhead and rework.
- 3Verify unique outcomesCount each fully resolved customer issue once after the check window.
- 4Segment comparable workRead similar issue groups before comparing channels or teams.
- 5Protect customer outcomesKeep reopens, repeat contact, effort and quality beside the cost.
That distinction matters in messaging support. One customer issue may create a WhatsApp thread, a follow-up on another channel and several internal tasks. Cost per contact prices each interaction. Cost per resolution asks what the business spent to make the underlying problem go away.
Choose one resolution unit before calculating anything. It could be an order problem, access issue, billing dispute or other customer need. Write the rule beside your conversation closure criteria so analysts and team leads apply the same definition.
Build the Whole Cost Pool
Start with a complete period such as one month. Include direct agent pay and employer costs, then add the work that makes those agents effective. Team leads, quality reviewers, workforce planners, training, knowledge maintenance, software, telephony, messaging fees, equipment and a reasonable share of overhead all support the resolution.
Rework belongs in the cost pool too. A repeat contact, avoidable escalation or reopened case consumes real time even when another department absorbs it. The failure demand method helps identify work created by an earlier service failure without calling every repeat contact waste.
| Cost input | Include in the shared pool | Allocate with care |
|---|---|---|
| Direct labour | Pay, benefits, payroll costs and overtime | Use paid support time, not only chat time |
| Indirect labour | Team leads, quality, planning and operations | Share cost across the work they support |
| Technology | Inbox, AI, telephony, messaging, knowledge and analytics | Include implementation and monitoring where material |
| Enablement | Hiring, onboarding, coaching and documentation | Spread one-time costs over a stated period |
| Overhead | Equipment, facilities, security and shared services | Use a consistent finance-approved rule |
| Rework | Reopens, repeat contacts and avoidable escalations | Attribute it to the originating issue where possible |
Do not hide a disputed allocation inside the final number. Publish the cost map with the result. Finance can then challenge one assumption without dismissing the whole measure.
Count Verified Resolutions Once
Build the denominator from a fixed cohort. Take issues first marked resolved during the period, give every issue the same observation window and remove anything that did not meet the definition. Count each unique issue once, even if it required several conversations.
Suppose a hypothetical support operation spends $82,000 in a month. It closes 7,100 records, but 300 are duplicates, 180 are abandoned without an outcome and 220 reopen inside the agreed window. The verified denominator is 6,400 unique resolutions. Cost per resolution is $12.81, not $11.55.
The exact reopen window depends on the work. A password reset may prove itself quickly, while a delivery correction needs time to reach the customer. Keep the window stable within an issue group and track the customer support reopen rate beside the cost measure.
Also look for the same issue returning in a new conversation. Status history alone can miss that. Customer identity, issue category and outcome evidence should connect the records without merging genuinely new needs.
Segment Before You Judge
A blended average can be mathematically correct and operationally useless. A simple address change should not be compared with an account recovery that requires security review. Segment by issue type first, then by channel, language, customer tier or automation path only when the groups remain large enough to interpret.
Use the same cost-allocation rule in every segment. If specialist engineering time appears only in one team’s cost but shared platform expenses disappear from another, the comparison rewards accounting choices rather than better service.
Compare a segment with its own earlier periods before comparing agents or teams. The number is designed for process decisions, not a leaderboard.
Read Cost Beside Customer Outcomes
A falling cost is not automatically an improvement. It may mean better knowledge, routing or automation. It can also mean agents are closing early, customers are giving up or work is shifting to another queue.
Read cost per resolution with reopen rate, repeat contact, customer effort and quality review. The observed customer effort method keeps repeated information, transfers and channel changes visible. A conversation quality scorecard checks whether the answer was accurate, complete and safe.
Apply the same discipline to automation. Include licensing, usage, implementation, maintenance, monitoring and human escalation. Count an automated resolution only when the customer’s issue is complete. A cheap partial answer that creates a second contact has moved cost rather than removed it.
Use the Number to Change Work
When one issue group costs more, sample the underlying conversations. Look for missing knowledge, unclear product steps, weak intake, unnecessary approvals, repeated handoffs or failures outside support. The useful action may be a clearer policy or product fix rather than faster agent handling.
A shared event history makes the calculation easier. In DripTell’s team inbox, assignment, status, notes and conversation context can sit with the thread. The team still needs finance-approved cost rules and a verified-resolution definition. Software cannot decide those honestly on its own.
Recalculate on a stable cadence and preserve the old assumptions. A trend is useful only when a changed number reflects changed work rather than a changed denominator.
Frequently Asked Questions
What is the difference between cost per ticket and cost per resolution
Cost per ticket usually divides support cost by ticket records. Cost per resolution divides the complete cost pool by unique issues that were fully solved and remained solved through the chosen window. Several tickets may belong to one resolution.
Should software and manager time be included
Yes. Include material direct and indirect costs that enable support, then document how shared costs are allocated. Excluding management, quality, tools or training understates the real cost.
Can AI lower cost per resolution
It can when it completes suitable issues accurately with lower total operating effort. Include the full cost of the AI path and keep reopens, repeat contacts, escalations and quality visible before calling the change a saving.
DripTell Editorial
Practical guidance reviewed by the DripTell product and customer workflow team.
See how DripTell checks product claims, uses primary sources and handles corrections.
Editorial and source policy