Every outsourced support dashboard we inherit looks healthy on paper. Average handle time down. Tickets closed up. First response inside target. Then you talk to the customers and find people who contacted support three times about the same problem and gave up after the second. The report was green because it measured how fast the team typed, not whether anyone's issue got solved.
That gap is the whole problem with measuring outsourced support. When the agents are not your employees, the scoreboard is the only steering wheel you have. You cannot walk the floor or read the room. You see what the report shows you, and a vendor will optimise for exactly the numbers you put in front of them. Put the wrong numbers up, and you get a team that is fast, busy, and quietly losing your customers.
So the real question is not "what metrics exist." Every provider can recite the list. It is which numbers actually prove quality, which ones flatter the vendor, and how to read them together so a good-looking dashboard cannot hide a bad operation. Here is the set we run our own 24/7 first-line desk on, what each number hides when you read it alone, and the review cadence that keeps a vendor honest.
Activity metrics tell you the team is busy, not good
Every support metric falls into one of two buckets. Activity metrics measure output: tickets closed, calls answered, average handle time, agent occupancy. Outcome metrics measure whether the work was any good: did the issue get resolved, did the customer have to come back, were they satisfied. Most vendor reports are heavy on the first bucket and thin on the second, because the first bucket is the one the vendor controls directly.
That control is the catch. A team can close more tickets by closing them faster and worse. It can cut average handle time by rushing customers off the line. Almost every production number improves when you degrade the actual service, which is exactly why a report built on production numbers tells you nothing about quality.
Here is the trap in one move. Average handle time drops from nine minutes to six. That reads as a 33% efficiency win. But if first contact resolution fell from 75% to 60% in the same period, the team did not get more efficient. It learned to end conversations before the problem was solved, and those customers come back as second and third tickets you pay to handle again. The efficiency was an illusion, paid for in repeat contacts.
The six support metrics that actually prove quality
You do not need twenty metrics. You need six that, read together, are hard to fake. Each one proves something a customer would recognise, and each one hides something when you read it on its own.
| Metric | What it proves | Realistic target | What it hides on its own |
|---|---|---|---|
| First contact resolution | The customer got a complete answer the first time | 70-80% on first line | A high rate from agents closing tickets the customer reopens |
| Customer satisfaction | How the interaction felt to people who answered | Above 90% of surveyed tickets | The view of a small, self-selected few when response rate is low |
| Repeat-contact rate | Whether one problem becomes three tickets | Under one in five | Nothing, because most reports never track it |
| First response time | How long a customer waits to hear back | Chat under 2 minutes, email under a few hours | A fast holding reply with no resolution behind it |
| Resolution within target | Issues close inside the window you promised | 85% or more inside the commitment | A tidy median while the hard tickets quietly age |
| Quality assurance score | The answer was correct, compliant, and on tone | Agreed rubric, calibrated | Little - unless the vendor scores it alone and you never check |
For a first-line desk, realistic targets are not aspirational: first contact resolution in the 70 to 80% range, customer satisfaction above 90% on the tickets that get surveyed, repeat contacts under one in five, and the large majority of issues resolved inside the window you promised. The exact numbers vary by product and channel. What does not vary is the need to hold all six at once, because the moment you reward one in isolation, that is the one the vendor will move, at the expense of the others.
Every good-looking number can hide a bad one
The single most useful habit in reading a support report is to never read a metric alone. Read it against the number it can hide behind. The pairs are where the truth lives.
Handle time against first contact resolution: fast is only good if the problem stayed solved. Customer satisfaction against survey response rate: a 95% satisfaction score on a 4% response rate is the opinion of the few people happy enough to answer, not your customer base. Tickets closed against repeat-contact rate: a closed ticket that reopens tomorrow was never resolved, just filed. First response time against resolution time: a vendor can hit a two-minute first response with an automated holding reply, then let the real work sit for three days. Each metric on its own can be staged. Read in pairs, they corroborate each other, or they expose the gap.
The survey score is not your quality score
Customer satisfaction is the metric everyone leans on, and it is the easiest to over-trust. It only captures customers who answer the survey, and it only captures how the interaction felt, not whether the answer was correct, complete, or compliant. On a regulated or high-stakes desk, a confident wrong answer can score five stars and still be a problem you clean up later.
The control for that is a manual quality assurance review, scored by you, not only the vendor. You sample a fixed share of tickets each week, score them against a written rubric (was the issue resolved, was the answer accurate, was policy followed, was the tone right), and then you calibrate. Calibration means you and the vendor score the same tickets independently and reconcile where you disagree, so that an eight-out-of-ten ticket means the same thing on both sides of the contract. Without calibration, the vendor is grading its own homework.
This is where running the desk ourselves changes how we measure it. On our own operation, we run 24/7 first-line support for an iGaming client inside their own Jira and client area, and the customer-survey number is the last thing we look at, not the first. We sample tickets every week, score them against a rubric the client signed off on, and calibrate the scoring with them directly. That calibration is what makes an outsourced scorecard trustworthy. It is also the step almost no vendor offers unless the client asks, because it is the one that takes away the vendor's ability to mark itself.
Put the targets in the contract, then review them monthly
Numbers only protect you when they are written down as commitments and reviewed on a cadence. A response-time commitment, a minimum first contact resolution rate, a customer satisfaction floor, and a quality assurance threshold belong in the agreement, with a defined trigger for what happens when a number is missed two periods running. The trigger should be a root-cause and remediation plan, not a service credit you will never get around to claiming.
Then the monthly review has to do more than display green numbers. A review worth having interrogates four things:
- Which metric pairs diverged: handle time against resolution, closed against reopened
- Where the quality assurance calibration disagreed, and why
- The repeat-contact and escalation trend over the last three months, not just this one
- Every off-target number, who owns it, and the plan to move it
If the monthly review is a vendor presenting a wall of numbers it chose, you have outsourced the measurement along with the work. The point of the cadence is to keep the hard numbers visible before they turn into churn.
How IMMIDO measures support quality
We do not hand a client a green dashboard and call it quality. We track the paired metrics, run a weekly quality assurance calibration on real tickets, and use the monthly review to surface the numbers that are not moving the right way, not to hide them. Because the team works inside the client's own tools, the measurement runs on their data and their tickets, not a parallel system we control.
If you are running an outsourced support team and cannot tell from the report whether it is actually any good, that is the conversation worth having. Our customer support operation is built to be measured on outcomes, not activity. Book a call and get a quote → We will look at the numbers you already have, show you the pairs your current report is hiding, and tell you what a trustworthy scorecard for your desk looks like.