How to write a treatment plan with SMART goals

A treatment plan without clear goals is a list of good intentions. With SMART goals it becomes a map: you and your client know where you are heading, how to measure progress, and when to celebrate that something was achieved. This guide covers the plan’s structure, the origin (and fine print) of the acronym, how to anchor goals to validated instruments, a sample plan calibrated to the PHQ-9’s actual severity bands, and where all of it lives inside the clinical record.

Where SMART comes from, and its fine print

The acronym was not born in the clinic. It is attributed to George T. Doran, who proposed it in 1981 in Management Review as a formula for writing management objectives. And there is a detail almost nobody mentions: Doran’s original letters were not today’s. His formulation was specific, measurable, assignable (to someone), realistic, and time-bounded; the A of “achievable” and the R of “relevant” are the modern adaptation — which happens to be the useful one in a clinical context.

Knowing this has a practical consequence: SMART is not a canonical clinical standard with an official definition, but a writing tool. What makes it valuable in psychotherapy is not the acronym — it is that it forces every goal to be measurable, reviewable, and correctable.

The minimum structure of a treatment plan

Before the goals, the plan needs context. A useful structure has four levels:

Presenting problem. The difficulty that motivates treatment, in concrete terms: “anxiety episodes with avoidance of public spaces,” not “anxiety.”
Long-term goal. The broad outcome the client is after: recovering social and work functioning.
Objectives. Specific, measurable steps toward that goal — this is where SMART comes in.
Interventions. What you will do: techniques, frequency, modality.

The most common mistake is jumping straight to interventions. Without intermediate objectives, there is no way to know whether graded exposure or cognitive restructuring is actually working.

What SMART means in a clinical context

S — Specific. Which behavior or experience will change, and in what context. “Reduce avoidance” is vague; “go to the supermarket unaccompanied” is specific.
M — Measurable. With a number, a frequency, or a validated instrument: times per week, PHQ-9 score, minutes of tolerated exposure.
A — Achievable. Realistic for this client, at this point. A goal that gets missed three weeks in a row doesn’t motivate — it demoralizes.
R — Relevant. Connected to what matters to the client, in their own words. Goals the client doesn’t feel are theirs get abandoned.
T — Time-bound. A deadline for evaluating it. Not an expiration date: an appointment to review and adjust.

The numbers that make a goal measurable

M is the letter that separates a reviewable plan from a statement of intent, and the most solid way to honor it is to anchor goals to validated instruments. For depression, the PHQ-9 is the ideal example: nine items scored 0 to 3, for a total of 0 to 27. In the original validation study, scores of 5, 10, 15, and 20 mark the thresholds for mild (5-9), moderate (10-14), moderately severe (15-19), and severe (20-27) symptoms, and a score ≥ 10 showed 88% sensitivity and 88% specificity for major depression. There is a citable threshold for the far end of treatment too: the official CMS quality measure in the United States defines depression remission as a PHQ-9 below 5 points at twelve months, in patients who entered care scoring above 9. With those bands, “improve mood” becomes “get from 16 to under 10,” and remission stops being an impression and becomes a number.

Two practical advantages: the PHQ-9 and GAD-7 are free — the official site phqscreeners.com states that the instruments and their translations can be downloaded and reproduced with no permission required — and if you work with Spanish-speaking clients, the Spanish PHQ-9 has been validated at scale in Mexican samples. The usual caution applies: no score is a diagnosis. Instruments screen and monitor; diagnosis integrates your clinical judgment.

Before and after: three examples

Vague: “Improve mood.” SMART: “Complete at least 3 planned pleasant activities per week for the next 4 weeks, recorded in the behavioral activation log.”

Vague: “Work on social anxiety.” SMART: “Start a brief conversation with a coworker at least twice a week for 3 weeks, recording anxiety level (0-10) before and after.”

Vague: “Reduce couple arguments.” SMART: “Use the agreed pause (20-minute time-out) in at least 80% of escalating arguments over the next 6 weeks, per both partners’ logs.”

Notice what these goals have in common: they describe the client’s behavior, not the therapist’s. “Apply cognitive restructuring” is your intervention; the goal is what changes in the person’s life.

Measuring often is not paperwork: it is measurement-based care

The practice of administering scales systematically and using the results to make decisions has a name: measurement-based care. Scott and Lewis define it as the use of systematic data collection to monitor client progress and directly inform care decisions, and report that, used as a framework to guide practice, it results in superior client outcomes compared to usual care. In their case example, the measures are administered before each session, so the data point is already on the table when the session starts.

For the treatment plan, the consequence is direct: a goal anchored to an instrument only works if the instrument is administered regularly. A PHQ-9 every six months tells you nothing in time; the same instrument given session by session — it takes two or three minutes — turns stagnation or deterioration into a visible signal on the graph, weeks before it becomes obvious in the narrative. The plan sets the destination; frequent measurement tells you whether you are moving toward it. How to set up that session-by-session tracking — instruments, cadence, and thresholds — is covered in measuring therapy progress.

Goals agreed on, not assigned

A technically flawless goal the client does not feel is theirs gets abandoned. The framework for avoiding that is shared decision making: Slade defines it as the process in which clinician and patient work together to select treatments based on clinical evidence and the patient’s informed preferences. Honesty requires a caveat: the evidence on its effect on clinical outcomes in mental health remains inconclusive, so present it as good practice, not as a proven technique. But one figure Slade cites illustrates the size of the problem: in surveys of the English NHS, only 42% of patients fully agreed that they had agreed with services what care they would receive.

In practice: write the long-term goal in the client’s own words (in quotes if needed), negotiate each objective’s success criterion — “how would you and I know this got better?” — and give the client a copy of the plan. A plan the client can read and correct is also a natural extension of informed consent: they decide, with information, what treatment they are accepting.

A sample plan (moderately severe depression)

Presenting problem: moderately severe depressive symptoms (PHQ-9: 16, band 15-19) with social withdrawal and work absenteeism.

Long-term goal: “get back to functioning at work and with my people” (client’s words); in plan terms, recover day-to-day functioning and bring symptoms below the moderate range.

Objective 1: increase personally meaningful activities to 4 per week within 4 weeks (behavioral activation log). Objective 2: reduce the PHQ-9 from 16 to under 10 within 12 weeks, administering the instrument at each session; remission horizon: a score below 5. Objective 3: return to full work attendance within 6 weeks, starting at 3 days per week.

Interventions: weekly behavioral activation; cognitive restructuring from week 3; coordination with psychiatry for medication evaluation.

One detail of this plan worth copying: every severity label matches the instrument’s validated bands. A PHQ-9 of 16 is not “moderate depression” — in the original bands, 10-14 is moderate and 15-19 is moderately severe. Mislabeling severity is not cosmetic: it shifts the baseline against which you will judge every later gain.

Four frequent goal-writing mistakes

Labels that don’t match the score. If the plan says “moderate” and the instrument says 16, the first review is born confused. Use the published bands and cite them.
Goals without a baseline. “Reduce the PHQ-9 to under 10” can only be evaluated if the plan records where it started. No baseline, no delta.
Too many goals at once. Five or six simultaneous fronts dilute effort and guarantee the kind of misses that demoralize. Two or three active objectives are usually enough; the rest wait their turn.
Deadlines without a review appointment. A “12-week” goal with no scheduled date to review it is a decorative deadline. The T includes when it gets evaluated, and with whom.

Review the plan as part of treatment

A plan written once and filed away is useless. It helps to separate two rhythms. Measurement runs session by session, or at whatever pace you set for each instrument. The formal plan review is something else: sitting down with the client to compare goals against data and adjust. No clinical guideline dictates a cadence; in our experience, a practical convention that works for most outpatient cases is reviewing it every 4 to 8 weeks — and, in addition, whenever the scores show stagnation or deterioration, without waiting for the date. At that review, achieved goals get celebrated and replaced, stalled ones get recalibrated or broken into smaller steps, and goals that stopped being relevant get retired with the same transparency with which they were agreed.

Where the plan lives in the record

The treatment plan is not a loose document: it is a named component of the clinical record. The APA’s Record Keeping Guidelines — adopted in 2007 and still the most recent published edition, framed as aspirational guidance — list a “plan for services, updated as appropriate” among the contents of the psychological record, alongside informed consent, the presenting complaint, and per-contact notes. Note the wording: updated as appropriate. A plan frozen at intake does not meet its own definition. The plan and the session note also need each other: the P of each SOAP note is the next move within the map the plan draws, and notes that say “continue treatment” session after session almost always betray a plan with no verifiable goals behind it.

On retention, the same guidelines suggest — explicitly “in the absence of a superseding requirement” — keeping full records until 7 years after the last date of service for adults, or until 3 years after a minor reaches the age of majority, whichever is later. That is US guidance, and local law prevails where it exists: in Mexico, for example, the clinical-record norm (NOM-004-SSA3-2012) sets a minimum of 5 years from the last act of care. Wherever you practice, verify your jurisdiction’s rules and treat the legal period as a floor.

From the plan on paper to the plan you actually use

Writing a good plan takes time; keeping it measured and current takes more. gesell.ai generates treatment plans with SMART goals from your intake assessment — session-by-session roadmaps — and keeps the plan and your session notes in one structured chart; it also includes built-in validated scales with automatic scoring, such as the PHQ-9 and GAD-7. You set the clinical direction; the platform makes sure it is on paper.