The customer service metrics that matter (and how they get gamed)

Support teams drown in dashboards. Most metrics on them are either vanity numbers (total tickets handled) or averages that hide the truth (mean response time smoothed by a thousand easy chats). This guide covers the handful of customer service metrics that actually change decisions, what "good" roughly looks like, and the standard ways each one gets gamed, because every metric that becomes a target gets gamed.

First response time (FRT)

What it is: How long a customer waits for the first human (or automated) reply.

Why it matters: It is the loudest signal customers feel. A fast first response buys patience for everything after it; silence compounds frustration faster than any other failure.

Watch: Median and 90th percentile, per channel. Averages hide the disasters, and chat expectations (minutes) are nothing like email expectations (hours).

How it gets gamed: Auto-replies and "we received your message" acknowledgements that technically stop the clock while helping nobody. Count only substantive first responses. We wrote a full playbook on this in how to reduce first response time.

Resolution time

What it is: Time from first message to actually solved, not to first reply.

Why it matters: Customers remember how long the whole problem took, not how quickly you said hello.

Watch: Median per issue category. Password resets and billing disputes should not share a target; blended resolution time is close to meaningless.

How it gets gamed: Closing conversations prematurely and letting the customer reopen. Pair it with reopen rate to keep it honest.

CSAT (customer satisfaction)

What it is: The post-conversation "how did we do" rating, usually a 1 to 5 scale reported as the percentage of positive scores.

Why it matters: It is the most direct read on individual conversations, and trends per agent, per category, and per channel expose problems the volume numbers hide.

Watch: Response rate alongside the score. A 95% CSAT from 4% of customers mostly measures who bothers to click. Read the free-text comments; the number is the headline, the comments are the article.

How it gets gamed: Agents asking happy customers to rate and quietly not asking the angry ones. Automate the send on every closed conversation.

Reopen rate

What it is: The share of "solved" conversations the customer reopens.

Why it matters: It is the truth serum for both resolution time and CSAT. Low resolution time with high reopens means you are closing tickets, not solving problems.

Watch: Per agent and per category. A single agent with double the team's reopen rate is a coaching conversation; a single category with high reopens is usually a product or docs problem.

Self-service rate

What it is: The share of customer questions answered without a human: help-center reads that do not become conversations, plus AI-agent resolutions that never reach a person.

Why it matters: It is the only metric here that scales headcount-free. Every point of self-service is support capacity you did not have to hire.

Watch: AI resolution rate together with its handoff quality (do people who reach a human after the AI end up satisfied?), and top help-center searches with zero results, which are literally a content request queue. The craft side of this lives in writing help articles people actually find.

How it gets gamed: Making humans hard to reach and calling the resulting silence "deflection". Deflection that frustrates shows up later in churn, where it is far more expensive.

Volume per customer (ticket rate)

What it is: Conversations per 100 active customers per month.

Why it matters: Raw ticket volume rises with growth even when the product improves. Normalizing by customers separates "we are growing" from "the product is confusing people more than it used to".

Watch: The trend after each release. A jump tied to a launch is the fastest product feedback you will ever get.

What about NPS?

Net Promoter Score measures relationship sentiment, not support performance, and it moves with pricing, product, and brand as much as with your team. Useful company-wide; unfair as a support KPI. The differences deserve their own article: CSAT vs NPS, and when each one lies.

A starter dashboard that fits on one screen

1. Median and P90 first response time, per channel
2. Median resolution time, per category
3. CSAT with response rate
4. Reopen rate
5. Self-service rate (help center plus AI)
6. Conversations per 100 customers, trended

Six numbers, reviewed weekly, argued about honestly. That beats a forty-widget dashboard nobody opens. Chattypie's inbox tracks these out of the box, but the discipline matters more than the tool: pick the six, publish them to the team, and let every process change justify itself against them.