Most support dashboards measure what is easy to count rather than what matters. Tickets closed, messages sent, average handle time. All real numbers, none of which tell you whether customers are being helped.
Here are the metrics worth tracking, what each one actually tells you, and the ways each can be gamed without anyone intending to.
First response time
How long a customer waits before a human acknowledges them. This is the metric customers feel most directly, and the one most closely tied to whether a pre-purchase question turns into a sale.
Track the median and the worst ten percent, not the average. An average of four hours can hide a queue where most replies land in twenty minutes and a handful sit for two days. The people in that tail are the ones who write reviews about it.
How it gets gamed: automated acknowledgements. If a bot replies instantly, first response time collapses to seconds and means nothing. Measure time to a substantive human reply.
Resolution time
How long from first contact to the customer’s problem actually being solved. Where first response measures responsiveness, resolution measures competence.
Segment it by ticket type or the number is meaningless. A password reset and a disputed international return do not belong in the same average. Once segmented, resolution time tells you where your process is slow rather than just that it is slow.
How it gets gamed: closing tickets early. If closing fast is rewarded, tickets get marked resolved while the customer is still waiting, and reopen later as new tickets. Watch reopen rate alongside it.
First contact resolution
The share of issues solved in a single exchange, without a back and forth. This is the most underrated metric in support, because it captures something the others miss: whether your team has the information and authority to actually finish the job.
Low first contact resolution usually points at a structural problem, not a people problem. Agents lack system access, policy is ambiguous, or every refund needs approval from someone who is in meetings all day. Those are fixable, and fixing them improves nearly every other number at once.
Customer satisfaction
The direct question after an interaction: did we help. Useful, with two caveats that are easy to forget.
- Response rates are low and skewed. People who are delighted or furious respond. The satisfied middle mostly does not.
- It measures the interaction, not the outcome. An agent can be warm, prompt and clear while delivering an answer the customer hates, such as a policy that does not allow the return.
Read the written comments, not only the score. The comments tell you which policies are generating anger, which is information no score will give you.
Contact rate
Tickets per hundred orders. This is the one most teams never track, and arguably the most strategically useful.
Every other metric assumes tickets arrive and asks how well you handle them. Contact rate asks a better question: why are customers contacting you at all. A rising contact rate on flat orders means something upstream broke, unclear sizing, a confusing checkout, shipping notifications that stopped going out.
The best support improvement is often not a support change. It is a clearer product page that stops the question being asked.
What to do with these
- Pick three, not seven. First response time, first contact resolution and contact rate cover responsiveness, competence and root cause between them.
- Set a target you would defend to a customer. “Under four business hours” is a promise. “Improve response time” is not.
- Review the tail weekly. Your worst ten tickets teach you more than any average.
- Never tie individual bonuses to a single metric. Whichever one you pick will be optimised at the expense of the others.
The short version
Measure how fast you answer, how often you finish the job in one go, and how often customers need to contact you at all. Read the comments. Watch the tail rather than the average. Everything else is reporting for its own sake.
If you want a view of how your current numbers compare, we are happy to look at your queue and tell you which of these is your real constraint.
