Most support dashboards measure speed and volume because both are easy to count. Neither tells you whether the customer left with their problem solved, which is the only thing that matters.
Measure one: repeat contact rate
The share of customers who come back about the same issue within, say, seven days. This is the most honest quality signal available and almost nobody tracks it.
A team can look excellent on speed and closure volume while generating repeat contacts constantly, because fast, incomplete answers close conversations without solving problems. Repeat rate exposes exactly that.
Measure two: resolution within the first reply
What proportion of conversations are fully resolved by the first human response? High numbers mean agents have the information and authority to answer properly. Low numbers usually indicate a systems problem — agents cannot see order data, or cannot authorise a refund without approval — rather than a skill problem.
Read the conversations
Every metric here is a pointer. Twenty transcripts read properly each week will teach you more about your support quality than any dashboard, and it is the only way to notice tone problems, which no number captures.
Measure three: subsequent response time
Not the first reply — the gaps in the middle. Customers judge the whole conversation, and a fast opening followed by long silences is a common and badly-received pattern. Track the median gap between all replies, not just the first.
Measure four: block and complaint rate on support numbers
If people are blocking the number they use to get help, something is badly wrong — usually promotional messages arriving on a support thread, or automation people cannot escape. Watch it as a support metric, not only a marketing one.
Asking for satisfaction
Short surveys work on WhatsApp because replying is one tap. Keep it to a single question with three buttons, send it only after resolution, and never send it more than once per customer per week.
Two cautions. Response is voluntary, so results skew to the strongly satisfied and the strongly annoyed. And a survey sent after an unresolved conversation reads as tone-deaf, so gate it on actual closure.
Metrics that mislead
- Conversations handled per agent. Rewards closing fast, which is exactly the behaviour that raises repeat contacts.
- Average handling time. Some problems should take longer. Pushing this down pushes quality down with it.
- Containment rate in isolation. Improves if you hide the route to a human.
- Closure count. Measures activity, not outcomes.
A small, honest dashboard
Median and 90th-percentile first response, median subsequent response, repeat contact rate, first-reply resolution rate, and a weekly read of twenty transcripts. Five numbers and a habit. It fits on one screen, resists gaming better than most, and will tell you more than a dashboard with forty tiles.