Every support dashboard contains at least one number that is actively making the team worse. Usually it's the one leadership looks at first.
The problem isn't that the metrics are wrong. It's that most of them can be improved by doing something bad, and none of them mean anything alone.
The four that earn their place
Resolution rate. The share of conversations that ended with the customer's problem actually solved, no follow-up, no reopen. This is the closest thing support has to a scoreboard, and it's the hardest to fake, you can't improve it by replying faster or closing tickets early, because both create reopens.
Time to resolution. Not time to first reply. How long the customer waited between asking and being done. Measure the median and the 90th percentile; the average hides everything interesting.
Reopen rate. What share of closed conversations come back. This is your lie detector. If resolution rate is climbing and reopen rate is climbing with it, you aren't solving more, you're closing faster.
Contact rate per active account. Tickets divided by accounts, not raw volume. Raw volume goes up when the company grows, which tells you nothing. Contact rate going up means the product got harder or the docs got worse.
The three that quietly mislead
First response time. Enormously popular, almost meaningless on its own. An instant "Thanks, we're looking into it" scores perfectly and helps nobody. It's only useful paired with time to resolution.
Tickets closed per agent. This metric rewards the exact behaviour you don't want: short answers, premature closes, avoiding hard threads. If you must track throughput, pair it with reopen rate per agent or don't track it at all.
Raw CSAT. The response rate on satisfaction surveys is usually under fifteen percent, and skews to the delighted and the furious. Useful as a trend over a large sample. Dangerous as a judgement on any individual conversation.
Read them in pairs
Single metrics lie. Pairs are much harder to fool:
- Resolution rate with reopen rate, are we solving, or just closing?
- Time to resolution with first response time, are we fast, or just fast to acknowledge?
- Contact rate with release notes, did last month's launch create work?
- Automation rate with escalation quality, is the AI handling the easy half or the wrong half?
A metric you can improve by doing something you'd be embarrassed to explain is not a metric. It's a target.
What to do about automation
Once part of your volume is handled automatically, one more pair matters: how much is deflected versus how much is genuinely resolved. Deflection means the customer went away. Resolution means the customer's problem went away. These are very different outcomes that look identical in a ticket count.
Track them separately. If automated resolution is rising while escalations get harder on average, that's the healthy pattern, the routine work is going, and what reaches your team is the work that actually needs a person.
A dashboard worth keeping
If you strip it back, four numbers and their partners:
- Resolution rate, next to reopen rate
- Median and p90 time to resolution
- Contact rate per active account
- Automated resolution rate, next to escalation volume
Everything else is diagnostic, useful when you're investigating something specific, misleading when it's on permanent display.