Short answer
The core customer service metrics are first response time, resolution time, first contact resolution, CSAT and ticket volume, with reopen rate as the honesty check. Once an AI agent answers part of the queue, add AI resolution rate and track satisfaction for AI and human replies separately.- Speed
- First response time, resolution time, speed of answer, SLA compliance.
- Quality
- CSAT, first contact resolution, reopen rate, customer effort score.
- Volume
- Ticket volume, contacts per order, backlog.
- Efficiency
- Handle time, tickets solved per agent.
- AI agent
- AI resolution rate, handoff rate, draft acceptance.
- Average with
- The median for time metrics, so a few slow tickets don't skew it.
Most support dashboards measure too much and decide too little. Pick a few metrics from each group below, set targets on four to six of them, and read each one next to the metric that catches its trap.
Speed
How long customers wait.
- First response time: time from the customer's first message to the first human or AI reply. Report the median.
- Average resolution time: time from the first message to the problem being solved, not just the ticket closed.
- Average speed of answer: for chat and phone, how long until someone picks up.
- SLA compliance: the share of conversations answered and resolved within your targets. The SLA calculator works out the rate and when a reply is due.
The trap: speed metrics reward fast, incomplete answers. Read them with reopen rate.
Quality
Whether the customer's problem was actually solved.
- CSAT: satisfied survey answers ÷ all answers × 100, usually 4 or 5 on a 5-point scale.
- First contact resolution: issues resolved in one interaction ÷ all issues × 100.
- Ticket reopen rate: reopened tickets ÷ solved tickets × 100. The hardest metric to fake.
- Customer effort score: how easy customers say it was to get help.
The trap: CSAT hears only from customers who answer the survey. Report the number of responses beside it.
Volume
How much demand arrives and why.
- Ticket volume: conversations per day or week, split by topic and channel.
- Contacts per order: support conversations ÷ orders. For a store, the fairest way to compare months with different sales.
- Backlog: open conversations older than your response target.
The trap: total volume hides the mix. Break it down by request type, such as where is my order, returns and damaged items.
Efficiency
How much work the team gets through.
- Average handle time: (handling time + wrap-up) ÷ conversations handled. The AHT calculator turns it into daily capacity.
- Tickets solved per agent: solved conversations ÷ agents, per day or week.
The trap: push either one as a target and agents close conversations faster instead of resolving them. Use them for staffing, not for judging people.
AI agent metrics
Once an AI agent answers customers, measure it on its own.
- AI resolution rate: conversations the AI resolved without a person ÷ conversations it handled.
- Handoff rate and reasons: how often it passes to the team, and why.
- Draft acceptance: how often the team sends the AI's drafts as written or with small edits.
Skip deflection rate as a target: it rewards keeping customers away, not solving their problem.
Tracking them in Aide
Aide Analytics separates conversations the Aide AI Agent resolved on its own from those a person finished, shows CSAT for Aide and for your team separately, and lists handoff reasons by day and channel. Every number opens the exact conversations it counts, and Aide Helpdesk tracks first response, next response and resolution times against your SLA policies.