The customer service metrics worth measuring are first response time, resolution rate, customer satisfaction (CSAT), cost per resolution and channel coverage. The first three describe how well a team serves its customers, and they exist only in that team’s own ticket data. The last two describe what the software costs and where customers can reach the team, and those can be measured from outside, from published prices and published channel lists.
Queue Bench ranks the outside metrics on dated leaderboards: seat cost at 5, 25 and 100 agents, AI price per outcome and channel and language coverage. The inside metrics get definitions on this page and no ranking, because no outside party can measure them honestly.
The five metrics at a glance
| Metric | What it measures | Measurable from outside? | Where the number lives |
|---|---|---|---|
| First response time | The wait from a customer’s first message to the first reply | No | Your helpdesk’s ticket timestamps |
| Resolution rate | The share of conversations that end with the problem solved | No; vendor rates are claims | Your tickets, under one fixed definition |
| CSAT | The share of surveyed customers who rate the help as good | No | Your survey responses |
| Cost per resolution | Total support spend divided by resolved conversations | Partly: list prices are published | Your invoices, plus the seat cost and AI price leaderboards |
| Coverage | The channels and AI languages the software supports | Yes | The channel coverage leaderboard |
First response time
First response time is the wait between a customer’s first message and the first reply written for that customer. Report it as a median per channel, in business hours, with automatic acknowledgements left out, because an email confirming that a message arrived answers nothing.
AI agents change the number. An AI reply arrives in seconds, so a team that counts it sees response time fall overnight while the wait for a person stays the same. Splitting first response time into a first AI reply and a first human reply keeps both figures honest.
Resolution rate
Resolution rate is the share of conversations that end with the customer’s problem solved. AI agent vendors quote it often, and each vendor counts a resolution its own way.
Intercom bills a Fin outcome when the customer asks for no further help after Fin’s last answer, and also for a Procedure handoff, a disqualification or a self-serve routing. HubSpot counts a resolution once a conversation goes 72 hours without a handoff to a human. Help Scout counts one only if the customer does not escalate, search the knowledge base, ask more questions or say they need more help. The same conversation can be resolved under one rule and unresolved under another, so a resolution rate only compares across teams that share one definition.
CSAT
Customer satisfaction (CSAT) is the share of customers who rate a conversation as good on a short survey sent after it closes. The score depends on who receives the survey, when it goes out and how the scale is worded. The customers who answer may not be typical of everyone helped, so the score works better as a trend than as a level.
A CSAT figure from a vendor describes that vendor’s customers under that vendor’s survey. Track your own CSAT against your past months, split by channel, and read the comments next to the score.
Cost per resolution
Cost per resolution is total support spend in a period divided by the conversations resolved in it. Spend has three parts: agent seats, AI charges, and everything else (add-ons, telephony, staff time). The first two have published list prices.
Seat cost varies widely at the same team size. Among the ranked vendors at 25 agents on annual billing, Jira Service Management Standard comes in at $5,650 a year and HubSpot Service Hub Professional at $27,000. The seat cost run prints the plan, the billing mode and any seat-cap substitution behind each figure.
AI charges scale with volume. At 2,000 AI-resolved conversations a month, the published rates put the AI bill at up to $1,000 on HubSpot before its included credits, $1,500 on Help Scout, $1,980 on Intercom and $3,000 on Zendesk at its committed rate, all before seats. Each of those four bills at most one unit per conversation, so the figures are upper bounds at list price. The AI price per outcome run carries the 500 and 10,000 conversation bills and the unit definition for each vendor.
Your real cost per resolution also includes the conversations people resolve, which no price list shows. Divide the full monthly spend by every resolved conversation, human and AI, and track the figure over time.
Coverage
Coverage is where the software can reach customers: the channels it runs natively and the languages its AI agent handles. The coverage leaderboard scores 12 channels: email, live chat, voice, SMS, WhatsApp, Facebook Messenger, Instagram, X, an in-app SDK, a help center, a community forum, and Slack or Teams. A channel scores only when it is part of a plan; add-ons and third-party integrations are listed and do not score.
Crisp ranks 1st with 11 native channels of 12, and Zendesk and Zoho Desk share 2nd with 9. AI language counts are vendor statements and do not score: Zendesk states 80+ languages and Freshdesk 60+. The coverage run holds the quote and source for every channel cell.
What only your own data shows
- How fast your team replies, per channel and per priority.
- How many conversations end resolved, and how many reopen.
- How customers rate the help they got, and why.
- What a resolution costs once human time is counted.
No outside benchmark can see these, and Queue Bench publishes no rankings for them. Every number the site does publish follows one set of testing rules: a dated run, a source URL, and no ranking across different units.
Using the metrics to pick software
The outside metrics narrow a shortlist and the inside metrics judge a trial. Price and coverage set the shortlist for help desk software, customer service software and AI customer service software. Response time, resolution rate and CSAT, measured on your own tickets during a trial under your own definitions, decide between the finalists. All three launch metrics sit together on the leaderboards index.