support cost per ticket: how to compute it

TL;DR

Cost per ticket is total support cost divided by tickets handled, but "total support cost" has to be fully loaded: salaries plus benefits, tools, overhead, and a share of management. Compute it monthly, segment by channel, and remember that deflecting cheap tickets raises the average. That is fine. It means the mix got harder, not that you got worse.

The query

support cost per ticket: how to compute it

Use this when

• Planning next year's support budget
• Deciding whether to outsource
• Proving an efficiency project paid off
• Leadership asks what support costs

Not for

• Tracking cost per individual agent
• Real-time dashboards
• Comparing support ROI to sales ROI
• Blaming a team for a high number

Steps

1. Add up fully-loaded cost

Salaries plus benefits and payroll taxes, ticketing and tooling licenses, office or remote stipends, and a fair share of management and QA time. If finance has a loaded-cost multiplier, use it. If not, 1.25 to 1.4 times salary is a defensible estimate. Leaving out overhead just moves the argument to later.

Expected output: one monthly total cost number everyone agrees includes everything.

2. Pick the period and stick to it

Monthly is the standard. Shorter windows swing with holidays and launches; longer windows hide trends. Compute it the same way every month or the trend is meaningless.

Expected output: a fixed monthly cadence with a frozen method.

3. Divide by tickets handled

Tickets solved plus tickets closed, per the period. Decide once whether reopened tickets count once or twice, and whether automated resolutions count at all. The denominator definition matters as much as the cost side.

Expected output: cost per ticket for the period, computed consistently.

4. Segment by channel

Chat, email, and phone have wildly different costs per ticket. A blended number lets a cheap channel subsidize an expensive one in the reporting. Segment first, blend second, and put automation savings where they belong: usually chat and self-serve.

Expected output: per-channel costs with the blend shown last.

5. Read the trend, not the number

No absolute number is good or bad without context: ticket complexity, channel mix, and labor market all move it. What matters is the direction after a change: did the knowledge base launch bend the curve, did the new tooling pay for itself.

Expected output: a trend line annotated with the changes you made.

Variant phrasings

how to calculate cost per support ticket

The full five steps. Calculation questions are usually step 1 and 3; the rest is interpretation.

customer support cost per contact

Same method, but "contact" includes chats and calls that never became tickets. Define the denominator to match how your team counts work.

support efficiency metrics

Cost per ticket plus handle time plus deflection rate. Efficiency is the trio, not any single number.

Why it happens

Leadership asks for cost per ticket because support looks like a cost center and cost centers get divided by output. The number is useful for budgeting and for proving that investments pay off, but it punishes success in one specific way: every ticket you deflect is usually a cheap, easy one, so the remaining mix is harder and the average rises. Teams that do not understand this stop deflecting to protect the metric, which is exactly backwards.

Edge cases

• Outsourced vs in-house mix: compute separately. Blending hides which one is actually cheaper.
• Shared agents across support and success: allocate by time split, honestly estimated. Guessing 50/50 every month is fine if you say so.
• One huge enterprise ticket: a single 40-hour escalation can wreck a small team's month. Note outliers; consider a trimmed average.
• Currency and region differences: a global team needs per-region costs before any blended number means anything.
• Leadership compares you to a benchmark: ask for their denominator definition first. Half of all benchmark arguments are definition mismatches.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstQ5A8zm_UOruOSY9XHmrug