A support dashboard needs six numbers: cost per resolution, genuine resolution rate, reopen rate, first response time at the ninetieth percentile, satisfaction segmented by topic, and arrivals versus resolutions. Leave off deflection, containment, and tickets closed per agent.
Key takeaways
- Report medians and percentiles, not means. The mean hides the tail.
- Reopen rate is what makes every other number trustworthy.
- Arrivals versus resolutions predicts the backlog. Queue length only describes it.
- Metrics that count abandonment as success will eventually embarrass you.
Most support dashboards report activity. The useful ones report outcomes and can survive a sceptical question from someone in finance. Here are six numbers worth having and four worth removing.
The six to report
1. Cost per resolution
Total fully loaded support cost divided by issues genuinely resolved. Include salaries with employer costs, all tooling, AI spend, quality assurance, and management time.
This is the headline because it captures spend and outcome in one ratio, which makes it the only number that compares hiring against automation on the same basis. We wrote up how to build the model.
2. Genuine resolution rate
The share of contacts that ended with a correct answer and no repeat contact within a set window. State the definition on the dashboard itself, because a resolution rate without a definition is not a number.
Exclude abandonment and handoffs explicitly.
3. Reopen rate
The share of resolved conversations that come back, matched across channels rather than by ticket thread.
This is the metric that makes the other five trustworthy. Without it, resolution rate, cost per resolution, and any automation claim can all be improved by loosening what counts as done.
Most teams do not track it. Adding it is the single highest-return change you can make to a support dashboard.
4. First response time, median and ninetieth percentile
Never the mean. Response time distributions are heavily skewed, so the average is dragged by outliers and hides the tail where your worst experiences live.
Segment by channel. Chat expectations are measured in seconds and email in hours, and a blended number serves neither.
5. Satisfaction, segmented
By topic, by channel, and by whether the issue was resolved or escalated. Report the response rate alongside it, because a score from a self-selected minority needs that context.
The aggregate tells you there is a problem. The segments tell you where.
6. Arrivals versus resolutions
Two lines on one chart, daily. Their gap is the only number that predicts your backlog.
Queue length tells you where you are. This tells you where you are going, and it settles arguments about effort immediately because the arithmetic is either working or it is not.
The four to remove
Deflection rate. Counts customers who gave up as successes. Any programme reported on it will eventually be contradicted by churn data.
Containment rate. Same problem. It measures where a conversation ended, not whether the customer was helped, and it can be improved by hiding the human option.
Tickets closed per agent. Rewards closing rather than solving. Individual targets on this reliably raise reopen rate and total cost.
Average handle time as a headline. Useful for capacity planning within an issue type, misleading everywhere else. It rises when automation removes the short contacts, which looks like a regression and is arithmetic.
What changes when automation arrives
Three things move, and one of them will alarm someone if you have not warned them:
Human cost per contact rises. Automation removes the cheapest contacts first, so the remaining human mix is harder by definition. Explain this before it happens.
Total contacts rise while tickets fall. More customers get answers, fewer become tickets. Report both or the picture is misleading in either direction.
Resolution rate becomes the number to scrutinise. Especially if you are billed on it. Ask for audit rights and use them quarterly.
Presenting it
Six numbers, one page, with the definitions written on it.
Add one sentence of narrative: what changed, why, and what you are doing about it. A dashboard without that sentence invites people to invent their own explanation, and their explanation is usually worse than yours.
The credibility test
Before adding any metric, ask: could this improve while the customer experience gets worse?
Deflection can. Containment can. Tickets closed can. Handle time can.
Cost per resolution paired with reopen rate and satisfaction cannot, which is why those three belong together and why a dashboard built on them survives scrutiny that a dashboard built on activity does not.
Frequently asked questions
What support metrics should I report to leadership?
Why should I not report deflection?
Should I report tickets closed per agent?
Why report the ninetieth percentile rather than the average?
Resolve, don't deflect.
See Fidiora resolve a ticket, capture a lead, and keep the bill predictable.