The test for including a metric

Ask one question of every number you are about to put in front of a client: if this moved twenty percent, would we do something differently? If the answer is no, it is context at best and filler at worst.

Most MSP dashboards fail this test cheerfully. They report everything the tooling can count, which is not the same as everything worth deciding on.

The eight that usually pass

These are aggregate, comparable across a quarter, and available from a normal PSA or RMM export without opening anything sensitive.

  • Tickets opened — demand direction. It cannot tell you severity or cause.
  • Tickets closed — throughput against that demand. Closure without opened counts means nothing.
  • SLA breaches — the promise you are measured on. Absolute counts matter more than percentages at small volumes.
  • Patch compliance — exposure you control. A number without a target is a fact, not a decision.
  • Backup success rate — the recovery position. Anything under complete deserves a named owner.
  • Critical alerts — noise or genuine risk, and worth splitting before the meeting.
  • Recurring incidents — the strongest signal that something structural is unresolved.
  • Unused licences — the one metric that reliably converts into money the client can see.

Why comparison beats history

One period is a number. Two comparable periods are a direction, and direction is what a client can act on. A quarter against the previous quarter answers whether the last decision worked, which is the only honest way to earn the next one.

Long histories look thorough and decide nothing. If a trend needs twelve months to be visible, it is not this quarter's decision.

Say what each number cannot prove

Ticket volume does not prove workload. Patch coverage does not prove security. Backup success does not prove recoverability, because a successful job and a tested restore are different claims.

Naming these boundaries in the room costs nothing and buys the credibility that makes the recommendations land. It also protects you later, when a client asks why an incident happened in a quarter that looked healthy.

The metrics that look useful and are not

Average resolution time is the most common offender. It mixes a five-minute password reset with a three-day escalation, and the average moves for reasons nobody in the room can act on. If you report it, report the distribution or do not report it.

Customer satisfaction scores collected from a handful of responses are the second. At small volumes the number swings on one bad week, and defending it costs more credibility than it earns. Uptime percentages are the third: they are true, they are close to a hundred, and they change no decision.

Setting a target before the next quarter

A metric without a target is a fact. A metric with an agreed target is a decision you can revisit. The agreement matters more than the number: a client who picks ninety-five percent patch coverage with you will defend it, and one who is handed ninety-eight will not.

Write the target, the owner and the date on the page you leave behind. Next quarter, the first line of the review is whether that target was met, which is the shortest path to a QBR that compounds instead of resetting.

Getting them out of your tooling

These eight fields, plus the period and a client pseudonym, are the entire input Sourcerow accepts. The narrowness is deliberate: a small, redacted contract is auditable, portable between PSAs, and safe to hand to a tool without granting access to anything.

The field-by-field definitions, with the validation each one is held to, are in the data dictionary.