What are SMART metrics?
SMART commonly stands for Specific, Measurable, Achievable, Relevant and Time-bound. It is a useful check on goals and targets. A metric itself also needs an operational definition: what is counted, for whom, over which period and using what source.
A goal, metric and target are different. “Improve access” is a goal. “Median waiting days” is a metric. “Reduce the median from 12 to 8 days by December” is a target. Do not treat the acronym as a substitute for a valid measure.
Apply the five checks
| Check | Ask… |
|---|---|
| Specific | What exactly changes, and for whom? |
| Measurable | Can the value be calculated consistently? |
| Achievable | Is the target plausible given the starting point and resources? |
| Relevant | Will the result inform the decision? |
| Time-bound | What period and review date apply? |
Achievable does not mean choosing an easy target to guarantee success. Explain the basis for the target and the conditions outside the team’s control.
Examples across workflows
| Vague statement | More useful specification |
|---|---|
| Improve training | Track the share meeting a defined skills rubric at completion, with baseline and coverage reported. |
| Speed up service | Track median days from a complete request to resolution, by month. |
| Improve partner reporting | Track reports received by the agreed deadline and the share passing defined quality checks. |
| Increase engagement | Name the behavior or scale, population and time window before setting a target. |
A metric dictionary template
| Field | Example |
|---|---|
| Name | Median request-to-resolution time. |
| Unit | Calendar days. |
| Start and end | Complete request received; resolution confirmed. |
| Population | Eligible requests closed in the reporting month. |
| Exclusions | Document agreed exclusions and report their count. |
| Source | Dated service records. |
| Owner | Service operations lead. |
| Review | Monthly, with unresolved requests shown separately. |
This definition still needs a check on unresolved requests: looking only at closed requests can hide a growing backlog. Pair the metric with a count and age of open requests.
Set a target after checking the baseline
Inspect the baseline distribution and variation before setting a target. Averages can hide different sites or groups. Consider capacity, seasonality and whether the measurement method will remain consistent.
If the definition changes, record a new version. A shorter measured time caused by changing the start event is not necessarily a real improvement. See baseline data.
Avoid incentives that distort the work
A narrow target can encourage behavior that improves the number while harming the purpose. Closing cases quickly may look efficient while leaving issues unresolved. Balance speed with quality and the person’s experience.
Ask what behavior the metric rewards and what it leaves invisible. Include exceptions and qualitative evidence in review rather than treating the dashboard as the whole picture.
Keep the definition with the result
Store the source, calculation, reporting period and version alongside the metric. When partners submit different measures, agree which are comparable and which should remain separate. AI can help identify inconsistencies, but a reviewer must resolve them.
Continue to the results framework guide to connect measures to intended change.
Frequently asked questions
Is SMART a measurement standard?
It is a planning heuristic, not a complete validation standard or universal reporting requirement.
Can a qualitative indicator be SMART?
It can have a clear definition, collection method and review date. Do not force it into a number that loses its meaning.
Should every metric have a target?
Not necessarily. Some measures describe context or emerging risks. Explain how each will be used.
Can targets change?
Yes, with a documented reason and approval process. Keep the original commitment visible.
