What are SMART metrics?
SMART metrics are measures that pass five checks: Specific, Measurable, Achievable, Relevant and Time-bound. The checks test whether a metric and its target are clear enough to act on; they do not, on their own, say how to count. SMART is a planning check, not a measurement standard.
Keep three things apart: “help people into good jobs” is a goal, and “placed in a job within 90 days of exit” is a metric. “Raise 90-day placements from [baseline] to [target] by year end” is a target, where SMART matters most.
THE SHORT VERSION
- Use the five SMART checks to make a goal and its target concrete, countable and dated.
- A metric can pass all five checks and still be counted two ways, so give each one a written definition row: who counts, what qualifies, when, from which source.
- Put SMART targets on outcomes measured per person, and keep output counts beside them to explain the result.
How do you apply the five SMART checks to a metric?
Ask one question per letter and answer it for your own metric, not in general. If any answer is “it depends”, the metric is not ready to report. The third column works one metric through the checks, using the fictional workforce fund from our free course.
| Check | Ask | “Placed in a job” (fictional) |
|---|---|---|
| Specific | What exactly changes, and for whom? | Completers start paid work |
| Measurable | Would two people calculate the same value? | Yes or no per person, from a placement survey |
| Achievable | Is the target plausible from the baseline and resources? | Set after last year's placements are counted |
| Relevant | Will the result inform a decision? | The fund reviews each partner on placement |
| Time-bound | What window and review date apply? | Within 90 days of exit, counted each reporting period |
Achievable does not mean a target you already know you will hit. Write down the basis for the number and what lies outside the team's control, such as local hiring.
Time-bound has two clocks: the window each person is measured over, such as 90 days from exit, and the period in which results are counted, such as a quarter. Mixing them causes many funder and partner disputes.
What do SMART metric examples look like, weak and strong?
A weak metric names an activity or a feeling; a SMART rewrite names the people, the change, the window and the source. Each rewrite below still leaves something to define, which is the subject of the next section.
| Vague statement | SMART rewrite | Still needs defining |
|---|---|---|
| Improve training | Share of completers meeting a defined skills rubric at exit, each quarter | Which rubric, who assesses |
| Speed up service | Median days from a complete request to resolution, by month | How open requests are shown |
| Improve partner reporting | Reports received by the agreed deadline that pass the agreed checks, per quarter | What “pass” means |
| Increase engagement | Name the behavior, the people and the window before setting a target | Everything: not yet a metric |
| Reach more people | Unique trainees enrolled in the period, not sessions, reported quarterly | How a re-enrollment counts |
| Help people find jobs | Completers placed in a job within 90 days of exit | Paid work only? Part-time? |
| Keep people employed | Placed trainees in the same job at 12 months, reported annually | How unknowns are reported |
| Raise wages | Hourly starting wage at placement, by training track, annually | Never one average across tracks |
The last four rows come from the reporting agreement of the course's fictional regional workforce fund, which funds four job-training partners, A to D. Each metric was agreed on the onboarding call, with its window and rhythm, before anyone reported a number.
Why does a SMART metric still need a definition?
Because two partners can each report a SMART metric and still count different people. SMART checks the shape of a metric; a definition row fixes who is counted, what qualifies, when, from which source, and how missing answers are treated.
In the fictional fund, Partner C reported 55 people placed “within 6 months”. That passes all five checks, but the agreement said 90 days. So the fund adds A's 42, B's 31 and D's 27 into a total of 100 and holds C's 55 until C confirms its 90-day count.
Enrollment breaks the same way when a partner counts sessions instead of unique people. The fix is one row per metric in a shared data dictionary, from metric to dimension to standard to roll-up.
Each row also names its collection point, data type, disaggregation, source, missing-value rule and owner. The chapter A shared data dictionary: metric, dimension, standard shows how to write and test every entry.
The video is a Sopact walkthrough that builds an IRIS+-aligned dictionary for a job training and placement program on demo data. Watch each metric get a definition and a Five Dimensions tag before any number is reported, and a number flagged as missing when its source was not selected.
Should a SMART metric count outputs or outcomes?
Put SMART targets on outcomes where you can, and keep output counts beside them. An output such as “participants enrolled” can be perfectly SMART and still say nothing about what changed for anyone.
In the fictional program, training sessions and mentoring produce outputs (participants enrolled, completed training) that lead to outcomes: placed in a job within 90 days, retained at 12 months, starting wage by track. The long-term aim is living-wage jobs.
An outcome metric is SMART when it is measured per person over a stated window, with non-answers counted as unknown. Completion still matters: it shows whether fewer placements mean fewer finishers. See output vs outcome for the full distinction.
How do you set a SMART target without distorting the work?
Count the baseline first, set the target from it, and pair each target with a check on what it might push people to do. A target chosen before anyone has looked at last year's numbers is a guess with a date on it.
Look at the spread, not only the average: results differ by site, partner and track, which is why the agreement asks for starting wage by track. Consider capacity, seasonality and whether the counting method will hold all year.
A narrow target can help the number and hurt the purpose. A 90-day placement target alone could favor any quick job; the agreement pairs it with retention at 12 months, so placements that do not last show up.
If a definition changes, note it in a change log and give the new metric its own row; moving the start event is not an improvement. The baseline data guide covers keeping a series comparable.
What goes in a SMART metric template?
One row per metric, with the SMART answers and the definition fields side by side, so the target and the counting rule never travel apart. Copy the first two columns into a spreadsheet and replace the example with your own.
| Field | What to write | Placed in a job (fictional) |
|---|---|---|
| Metric | A short name people will search for | Placed in a job |
| Definition | Who counts, what qualifies, when | A completer who started paid work within 90 days of exit |
| Target | Baseline, target, date | [baseline] to [target] by end of fund year |
| Denominator | Who the rate is out of | Completers in the period |
| Collection point · source | Which form, when | Placement survey 90 days after exit, linked to enrollment |
| Disaggregation | Groups to split by | Track, gender, age |
| Dimension · standard | Five Dimensions tag; code only if it fits | What · How much · no IRIS+ code |
| Missing-value rule | What a blank means | Unknown, never “not placed”; report the count |
| Roll-up | What may be added | Only counts that share the 90-day rule |
| Owner · review | Who approves changes, when reviewed | Fund portfolio manager · each reporting period |
An AI tool can run the checks and draft rows from your goals document. Treat the draft as a list of questions to settle.
PROMPT · PASTE INTO CLAUDE, CHATGPT OR YOUR AI TOOL
Below are our program goals and the metrics we plan to report to [FUNDER]. For each metric: 1. Run the five SMART checks. For each letter, say pass or fail and why. 2. Draft one definition row: metric, plain definition (who counts, what qualifies, when), denominator, collection point, source, disaggregation, missing-value rule, owner, roll-up rule. 3. Say whether it is an output or an outcome. Rules: - Do not invent numbers, baselines or targets. - If something is not stated, write "not in our data" and add a question. - A missing answer is unknown, never zero or "no". [PASTE GOALS AND DRAFT METRICS]
What can a SMART metric not tell you?
A SMART metric tells you whether a target was met, not why, and not whether the program caused it. Placements can rise because local hiring picked up. A before-and-after change needs a comparison, such as earlier cohorts or an outside figure on the same definition, before anyone calls it an effect.
Outcome metrics lean on follow-up answers, which are self-reported and incomplete. People who skip the placement survey may differ from those who answer, so report the unknowns next to the count.
AI can draft rows and flag vague definitions; people decide what a metric means and check drafted numbers against the record. In Sopact Sense, the AI Assistant answers only from the surveys you select, and each line links to a record you can open.
Start with one SMART metric this reporting cycle
Take the metric your funder reads first from goal to tested definition before touching the rest.
- Write the goal, the metric and the target as three separate sentences.
- Run the five checks, including both clocks for Time-bound.
- Fill one template row, with the denominator, the missing-value rule and an owner.
- Give a colleague five records, including a late answer and a blank follow-up; compare counts.
- Count the baseline, then set the target number from it.
- Pair the target with one check on what it might reward, such as retention beside placement.
After the first cycle you have one metric with a dated target, a definition two people apply alike, a baseline and questions for partners.
Frequently asked questions
What does SMART stand for in metrics?
SMART stands for Specific, Measurable, Achievable, Relevant and Time-bound. Applied to a metric, it asks what changes and for whom, whether two people would get the same value, whether the target is plausible, whether the result informs a decision, and which window applies. A written definition row is still needed so every team counts it the same way.
Is SMART a measurement standard?
No. SMART is a planning check, not a validation standard or a reporting requirement. It turns a goal into something countable and dated, but does not say who is counted, which source a value comes from or how missing answers are treated. Standards such as IRIS+ give metric definitions and codes; your data dictionary records how your metric relates to them.
Can a qualitative indicator be SMART?
Yes. A qualitative indicator can have a clear definition, a collection point, a window and a review date, such as “participants describe what got in the way of starting work, asked at the 90-day follow-up”. Do not force it into a number that loses its meaning. Report how many people answered, and read the themes next to the placement count.
Should every metric have a target?
Not necessarily. Some measures describe context, such as how many people enrolled, or watch a risk, such as placements that do not last a year. Give a target only where a decision depends on it and a baseline exists. For the rest, write down how the number will be used.
Can SMART targets change during a grant?
Yes, with a written reason and the approval of whoever owns the metric. Keep the original target visible beside the new one, so readers can see what moved and why. If the definition changes too, record it in a change log and treat the result as a new metric rather than overwriting the old series.
What is the difference between a SMART metric and a KPI?
A KPI is a metric an organization has chosen as one of its few key ones; SMART is a set of checks any metric or target can pass or fail. A KPI can be vague, and a SMART metric need not be key. Aim for a short list of key metrics, each SMART and each with a definition row every partner applies.
