The logic model and SMARTER objectives worksheet
What this is: A program design worked in the order that catches errors: the chain from inputs to impact with the assumption under each link, objectives written so they can fail, a capacity check run before launch, a measurement plan built at design time, and the standard that would mean the program should change or stop.
It assumes: A program you are designing, redesigning, or about to describe in a proposal. You do not need an evaluator, a data analyst, or a research budget.
Change before use: Every statement. The structure is the deliverable; the illustrations are invented and are marked for deletion.
This document explains what each part of a program design has to settle and how to work it. The worksheet itself is the accompanying spreadsheet — the logic model as one row per element with the assumption it depends on, the objectives with a column per SMARTER letter, the output-versus-outcome check, the implementation-capacity check, and the measurement plan.
The figures on the summary sheet are formulas. They count the links with no assumption stated, the objectives missing an element, and the outcomes that trace back to no activity. Read this once. Work in the spreadsheet.
The problem this fixes
Most programs in this sector are designed inside a proposal. The proposal is written under a deadline, by whoever writes well, and it is judged on whether it persuades. Nothing in that process tests whether the organization can operate what is being described. When the money arrives, the proposal narrative becomes the operating manual by default — and it was never built to be one.
The failure shows up later and looks like something else. Two staff describe the service differently. A funder asks for an outcome and the data collected cannot support the claim. The evaluation is designed in year two, so the baseline was never taken. None of these is a measurement problem. They are all the same design problem, arriving late.
This worksheet moves that work to the front, where it is cheap.
The chain, and the assumption under each link
A logic model is a chain of claims. Inputs make activities possible; activities produce outputs; outputs produce outcomes; outcomes accumulate into impact. Written that way it is nearly always true and nearly always useless, because the interesting part is not the chain — it is what has to be true for each link to hold.
So the worksheet asks for the assumption beside every link. Not a general statement of faith, but the specific condition that, if false, breaks that step.
| Link | The claim | The assumption that has to hold |
|---|---|---|
| Input → activity | Two full-time family specialists let us run weekly in-home sessions | We can recruit and retain two people with this qualification in this labor market at this pay rate |
| Activity → output | Weekly sessions produce twenty completed twelve-week cycles a year | Families attend at a rate high enough to complete; the cycle is not interrupted by placement changes |
| Output → outcome | Completed cycles produce improved caregiver capability | The curriculum addresses the actual driver of difficulty for these families, not a different one |
| Outcome → impact | Improved caregiver capability reduces disruption | Caregiver capability is a material cause of disruption in this population, relative to the other causes |
The assumption column is where a design gets honest. An assumption you cannot state is a link you have not thought about. An assumption you can state but cannot evidence is a risk you now know you are carrying, which is a better position than not knowing.
The two tests that catch most of it
- Every outcome traces back to an activity that could plausibly produce it. Read each outcome and name the activity. If the honest answer is "several things, generally," the outcome belongs to the organization's mission rather than to this program, and claiming it will not survive a serious reviewer.
- Every activity is one this organization can staff, supervise, document and evidence. Not could theoretically do — can do, at this size, with these people, alongside everything else already running. An activity that fails this test is a hiring plan or a systems build, and it belongs in the design as such rather than being assumed.
Most weak designs fail the first test at the top and the second at the bottom. The first failure produces a program that cannot demonstrate its effect. The second produces a program that cannot be delivered as described, which is worse, because it is discovered by the funder rather than by you.
Objectives that can fail
An objective stated as an activity — *provide services to a number of families* — cannot fall short of anything. It is a description of intent wearing the grammar of a target. If it cannot fail, it cannot teach you anything, and the review at the end of the year has nothing to review.
The worksheet uses SMARTER, with the last two letters doing real work:
| Letter | What it settles | The test |
|---|---|---|
| S — Specific | What changes, for whom, from what to what | Could two people read this and count the same thing? |
| M — Measurable | The number, the source, and the counting rule | Name the field in the system this comes from. |
| A — Achievable | Reachable with the resources actually committed | What is this based on — a prior period, a comparable program, or hope? |
| R — Relevant | Ties to an outcome in the logic model | Name the row in the model. If none, delete the objective or add the row. |
| T — Time-bound | The period, and the date it is measured | A date, not a season. |
| E — Evaluated | When this is examined, by whom, against what | A date in the calendar and a named person, before the period ends. |
| R — Revised | What happens when the evaluation says it is off track | The trigger and the action, written in advance. |
The distinction between the last two letters matters, and it is the same distinction this house makes in the corrective-action protocol. Reviewed is passive — a review can conclude that nothing needs to happen and still count as done. Revised commits to a change having been made, or to a written record of why none was warranted. A program that never revises an objective either got everything right the first time or never really looked.
Write the revision trigger before the program starts. Written afterward, it becomes an explanation.
Outputs and outcomes
The confusion is not carelessness. Outputs are what your systems already produce, so they are what gets reported; outcomes require a definition and a collection point that do not exist yet. The worksheet forces the pair.
| Output — what you did | Outcome — what changed | What the outcome needs before launch |
|---|---|---|
| Twenty families completed the twelve-week cycle | Caregivers report and demonstrate specific capability at exit that they did not have at intake | An intake measure, an exit measure, and the same instrument at both ends |
| Ninety staff hours of supervision delivered | Documentation defects in sampled records fall against the prior period | A sampling method and a rubric in use before the first supervision session |
| A resource library published | The people it was built for use it and can say what it changed | A definition of use, and a way to ask that does not only reach the enthusiastic |
The third column is the point of the exercise. An outcome without a pre-launch measurement plan is a claim you have decided to make later and will not be able to support.
The implementation-capacity check
Run this before the design is finished, not after it is funded. Each item is scored against the organization as it is today, not as the proposal describes it.
- Staffing. The roles exist, the qualifications are obtainable in this labor market at the rate budgeted, and the recruitment lead time is inside the launch schedule.
- Supervision. Someone qualified to supervise this work has the hours for it, and those hours are not already committed to something else.
- Documentation. The record each session or contact must produce has a place to live, and the person producing it has been shown what good looks like.
- Data. Every indicator in the measurement plan has a field, a collector, and a collection point inside the ordinary work rather than beside it.
- Regulatory. Any license, approval, or contract amendment the program requires has been identified, with the lead time stated. Confirm this against your own licensing authority and counsel.
- Finance. The unit cost has been built from the layers, not assumed from the award, and the gap — if there is one — is named.
- Space and systems. The physical and technical requirements exist or have a dated plan.
- Referral flow. There is a reason to believe the eligible population will arrive at the volume assumed, and a stated response if it does not.
A "No" here is not a reason to abandon the design. It is a dependency with a date, and it belongs in the launch plan. The failure this prevents is the one where the dependency is discovered in month two by the people delivering.
The revision point and the discontinuation standard
Two dates that almost no program design contains, and both of them protect you.
- The revision point. A date, set before launch, at which the first cohort's experience is read against the design and the design is changed. Without it the first cohort merely survives the program rather than improving it, and the second cohort gets the same thing.
- The discontinuation standard. What result, sustained over what period, would mean this program should change substantially or stop. Written at design time, by the people who believe in it, it is a statement of intellectual honesty. Written later, under pressure, it is a defense.
Boards and funders read a stated discontinuation standard as confidence rather than doubt. An organization willing to say in advance what failure would look like is telling you it expects to be able to tell the difference.
What this does not do
This worksheet designs a program and plans its measurement. It is not an evaluation, and completing it does not make a program evaluated. Where a funder or a regulator requires an independent evaluator, this is what you hand them so their work can start from a defined design rather than from reconstruction — and the evaluation itself belongs to someone with no stake in the result.
It also carries no requirement of your license, your contract, or your funder. Those are yours to confirm with your licensing authority, your counsel, and the award terms.
Free to adapt and use. No registration, no fee, no attribution required. This is a working template, not legal, tax, or accounting advice — the requirements that apply to your organization are confirmed by your own counsel, auditor, and licensing authority.