SMART metrics test whether a number is defensible: the five-test framework, six design principles, SMART indicators, examples, and FAQ for program teams.
SMART metrics are performance measures that pass five tests: Specific, Measurable, Achievable, Relevant, and Time-bound. The framework, first proposed by George T. Doran in 1981 for writing management objectives, turns a vague goal into a number a team can collect consistently, compare across periods, and defend under review.
The reframe this page argues for: a metric reports a number, while a SMART metric defends it. Program teams rarely lack metrics. What they lack is the metric that survives the board meeting, when someone asks what the number means, where it came from, and compared to what. A metric that passes four of five tests is precisely the one that breaks in that room.
Key takeaways
Compare two versions of the same claim. “We improved job outcomes for our graduates” reports an impression. “47 of 60 graduates placed in a training-matched role within six months, against a prior-cohort baseline of 38” defends itself: the denominator is stated, the window is stated, the comparison is stated. Every question a reviewer could ask is already answered inside the sentence.
Wording, though, is only half the discipline. The second sentence is checkable only if 60 graduate records exist, each carrying an enrollment date, a placement status, a role classification, and a link to the prior cohort's records. A SMART metric only becomes outcome evidence when each measurement lives on a participant record with a baseline and a follow-up. Sopact calls that record the Outcome Thread: one persistent record per participant, under a Contact ID assigned at first contact, where every wave of data lands. SMART wording without the thread is a well-phrased estimate.
Which numbers deserve this treatment is its own decision: counts of delivery are outputs, changes in people are outcomes, and the distinction is drawn carefully on the output vs outcome page. The infrastructure that keeps each metric attached to its records is the subject of outcome tracking software.
SMART does not measure performance. Applied to a performance indicator, SMART measures the quality of the indicator itself: whether the measure is precise enough (Specific), collectible as a number (Measurable), realistic to move (Achievable), connected to the outcome that matters (Relevant), and bounded by a deadline or window (Time-bound). A program can score badly on a well-built SMART metric; that is the metric doing its job.
The vocabulary around it sorts cleanly. A goal is the change you want. A metric is any quantified measure. An indicator is a metric chosen to stand for progress toward the goal. A KPI is an indicator someone has promised to be accountable for. SMART is the quality test you run on any of them, and a SMART indicator leans hardest on Relevant: it must evidence the outcome, and a proxy that is easy to collect but weakly connected fails the test no matter how precise it is. Training teams meet this exact trap constantly, which is why training metrics lean on completion rates that pass Measurable and fail Relevant.
Metric infrastructure evolved in three eras. The forms era made collection cheap: spreadsheets and survey builders gather numbers easily, but every collection is a disconnected snapshot, so nothing links this quarter's 68 percent to the people inside last quarter's 74. The dashboard era aggregated those snapshots into charts, which made the numbers presentable and left them exactly as defenseless: an aggregate cannot answer who, compared to what, or since when. The current era keeps every measurement attached to the person it describes, which is what makes a metric auditable.
The one evaluation test: ask the tool to show the people behind any reported number, each with a baseline and follow-up on one screen. If the answer is an export and a weekend of spreadsheet matching, the tool collects well and defends nothing. Teams that live this gap end up rebuilding the same metric by hand every reporting cycle, a pattern the impact measurement page traces across the whole evidence stack.
To write SMART metrics for a program, start from the outcome in your theory of change, draft the claim you want to be able to make, run it through the five tests, and lock the surviving wording in a data dictionary with a wave schedule. The sequence matters: teams that start from what is easy to collect end up with precise measures of the wrong thing.
Walk one metric through it. The program's theory of change says graduates gain stable, training-matched employment. Draft claim: “improve job outcomes.” Specific fails first: which jobs, which graduates? Revise to “graduates placed in training-matched roles.” Measurable asks for the number and the source: placements confirmed by follow-up survey, out of a stated denominator. Achievable calibrates the target against the prior cohort's 38 of 60. Relevant checks the measure against the outcome: placement within the field trained for, so any-job placements cannot inflate it. Time-bound closes the window: within six months of completion. The result is the metric from the opening: 47 of 60, training-matched, six months, against 38.
Then lock it. A data dictionary entry fixes the question wording, the scale, the denominator rule, and the wave schedule, so month nine measures the same thing as month one. The follow-up questions that populate the metric belong in the instrument from day one, drawn from sets like impact survey questions, because a Time-bound metric with no follow-up wave scheduled is a deadline with no data collection attached.
Workforce programs usually break on Measurable: the outcome, sustained employment, lives months after exit, so the metric depends on follow-up data the program has no mechanism to collect. Education programs break on Relevant: attendance and completion are effortless to count and weakly connected to learning, so the dashboard fills with proxies. Impact portfolios break on Specific: a portfolio-level metric like “lives improved” aggregates incomparable things, and diligence takes it apart in one question. Knowing which letter your program shape breaks on tells you where to spend design effort before the first number is ever collected.
Each letter of SMART catches a distinct failure mode, and a metric that passes four tests still fails review on the fifth. The map below is the checklist to run before a metric ships.
| Test | What it asks | How the metric fails without it |
|---|---|---|
| Specific | Exactly what is counted, for whom, under what definition? | “Improve outcomes” — every reader fills in a different meaning |
| Measurable | Can it be collected as a number, from a named source, repeatedly? | The outcome lives after exit and no follow-up mechanism exists |
| Achievable | Is the target calibrated to a baseline and real capacity? | An uncalibrated target reads as fiction and poisons honest reporting |
| Relevant | Does the measure evidence the outcome that matters? | Precise proxies: completion rates standing in for learning |
| Time-bound | By when, or within what window, is it measured? | A metric with no window can never be declared missed |
For standardized wording, the IRIS+ catalog offers field-tested metric definitions worth adapting before inventing your own. How each metric maps to the results chain is covered in the output vs outcome guide.
A metric reviewed quarterly can only tell you about a quarter that is already over. The point of building a defensible metric is to act on it, and that is the premise of the Loop, Sopact's method for continuous impact intelligence: collect clean at the source, analyze each measurement on arrival, improve while the cohort behind the number is still in the program. A SMART metric on the Loop is an instrument, and the KPI report becomes a byproduct.
The Loop also supplies the property SMART wording alone cannot: consistency of the measurement itself. A metric read across waves is only comparable if the same question, scale, and denominator hold every time, the same-answer-twice standard covered in Loop reliability. Clean-at-source collection enforces it at entry, which is cheaper than reconciling it at reporting time.
One method, three moves that never stop
Then the cycle runs again, a little sharper each time. Read the method: the Loop methodology →
The fastest way to find your weak letters is to run your current measures through the tests. Each prompt below is written to paste into Sopact Sense's Assistant, or to reason through with your team; the arrow above each one links the Academy walkthrough with the expected output and tips.
Academy walkthrough → How to build a data dictionary
Run each of these metrics through the five SMART tests: [PASTE METRIC LIST]. For every metric, name the first test it fails and rewrite it to pass all five. Then build a data dictionary entry for each survivor: exact question wording, scale, denominator rule, source, and wave schedule.
Academy walkthrough → How to build a logic model
Here is my program: [PROGRAM DESCRIPTION]. Place my current metrics on a logic model and classify each as an output metric or an outcome metric. For each outcome metric, state whether a baseline and a follow-up exist on the same participant record, and what it would take to make the metric defensible.
Academy walkthrough → Analyze pre, mid, and post survey data
Design the wave schedule that would populate these SMART metrics: [PASTE 3 METRICS]. Recommend the waves (intake, exit, 90-day, 180-day), the questions that must stay identical across every wave, and the denominator rule for reporting each metric when some participants miss a wave.
Academy walkthrough → The Loop methodology: continuous, not annual
My team reviews KPIs [CURRENT CADENCE, e.g. quarterly]. Using this cohort's data: [PASTE OR ATTACH], recompute my headline metric as of today and show which participants moved it since the last review, plus the open-ended answers from those participants that explain the movement.
Each walkthrough is practical and short: what to do, the prompt to run, the output to expect, and the tips that make it reliable.
Watch: turning program metrics into defensible outcome evidence with clean-at-source collection.
SMART metrics are performance measures that pass five tests: Specific, Measurable, Achievable, Relevant, and Time-bound. A metric reports a number; a SMART metric defends it under review. Sopact adds an infrastructure test to the wording test: the metric must live on a participant record, the Outcome Thread, with a baseline and a follow-up.
Specific (exactly what is counted, for whom), Measurable (collectible as a number from a named source), Achievable (calibrated to a baseline and real capacity), Relevant (evidencing the outcome that matters), and Time-bound (measured within a stated window). Sopact treats the five as failure tests: each letter names a distinct way a metric breaks in front of a board.
SMART measures the quality of the indicator, not the performance of the program. It tests whether a performance indicator is precise, collectible, realistic, connected to the outcome, and time-boxed. A program can score badly on a well-built SMART indicator; Sopact's view is that this is the indicator working, because an honest number you can act on beats a flattering one you cannot defend.
Vague: “improve job outcomes for graduates.” SMART: “47 of 60 graduates placed in a training-matched role within six months of completion, against a prior-cohort baseline of 38.” The denominator, window, definition, and comparison are all inside the sentence. In Sopact Sense, each of the 60 sits on an Outcome Thread, so the number is recomputed from records rather than asserted.
Start from the outcome in your theory of change, draft the claim you want to make, run it through the five tests in order, and lock the surviving wording in a data dictionary with a wave schedule. Sopact's sequence warns against the common shortcut: starting from what is easy to collect produces precise measures of the wrong thing.
A metric is any quantified measure. A KPI is a metric someone is accountable for. A SMART metric is either one that passes the five quality tests. Sopact's practical distinction: most dashboards are full of metrics, a few carry KPIs, and only the SMART ones with a baseline and follow-up on the same record survive a funder's diligence.
No. SMART governs the wording of a metric; outcome evidence needs infrastructure behind the wording: a persistent participant ID, a baseline captured at first contact, and follow-up waves on the same record. Sopact calls that record the Outcome Thread, and treats a SMART metric without one as a well-phrased estimate.
George T. Doran proposed S.M.A.R.T. in a 1981 Management Review paper as a way to write management objectives; it later spread from corporate goal-setting into program measurement. Sopact keeps Doran's original insight, that vague objectives cannot be managed, and extends it to evidence: vague metrics cannot be defended.
Continuously, not quarterly. A Time-bound metric reviewed after its window closes can only be reported, never acted on. Sopact's Loop methodology recomputes each metric as responses arrive, so a slipping number surfaces mid-cohort with the participant answers that explain it, while there is still time to intervene.
Next: see which of your numbers deserve the SMART treatment in output vs outcome, or how defensible metrics roll up into a funder-ready story on the social impact report page.
