What is the difference between an output and an outcome?
An output counts what a program delivered, such as people trained. An outcome is what changed for people afterwards, such as a job kept for a year. Outputs answer “did we do it?”; outcomes answer “did anything change for anyone?”
Both matter. Outputs show how the work was carried out, and they often explain an outcome that did not appear. Measuring an outcome still does not, by itself, prove that the program caused it.
THE SHORT VERSION
- Ask one question of any indicator: does it count something we did, or something that changed for a person?
- Word each outcome per person, with a window and a source, and keep people who did not answer as unknown.
- Report outputs beside outcomes, because delivery explains change, but never rename an output to make it sound like one.
Where do outputs and outcomes sit in the results chain?
Between activities and impact. Resources pay for activities, activities produce outputs, outputs lead to outcomes, and outcomes that last and spread become impact. The table follows the fictional workforce program from our free course, one of four job-training partners of a regional workforce fund.
| Level | Question | Workforce example |
|---|---|---|
| Input | What resources are available? | Trainer time, funding, practice equipment |
| Activity | What does the team do? | Training sessions and mentoring |
| Output | What was delivered? | Participants enrolled; completed training |
| Outcome | What changed for people? | Placed in a job within 90 days; retained at 12 months; starting wage by track |
| Impact | What longer-term change matters? | Living-wage jobs |
Read the arrows as claims. Completing training is expected to lead to a job within 90 days, and that job to last a year; each arrow is something the program believes, and the outcome indicators are how it checks.
Terms differ across funders and disciplines. The OECD evaluation glossary separates an intervention's products and services from its effects. Theory of change, logic model, logframe and results framework all use this chain under different labels; the chapter One change model, four formats shows the four side by side.
What are examples of outputs vs outcomes?
In every sector the output is the service delivered and the outcome is the condition that changed for someone afterwards. The last column is the question to ask before you claim the outcome.
| Setting | Output | Outcome | Evidence question |
|---|---|---|---|
| Training and workforce | People complete an assessment | People perform the task in practice | Was performance assessed the same way for everyone? |
| Customer experience | Support cases receive a response | Customers resolve the issue and keep using the service | Was resolution confirmed, or did the case only close? |
| Employee development | Managers complete coaching sessions | Employees see specific management behaviors change | Do follow-up accounts support the change? |
| Community services | Households receive an agreed service | Households can reach the support they need | Who is still underserved, and why? |
| Partner or supply chain | Suppliers submit quarterly evidence | The practice or condition improves | Was the condition verified, not only the form? |
A high output can be valuable and still not enough: a fast support response does not mean the problem was solved. A low output can explain a missing outcome, so read delivery and change together.
The short Sopact video below, Output vs Outcome: 7 Rules to Measure What Actually Changed, sets out rules for telling delivery from change. Watch it next to the six-rule table in the next section; both ask what changed for a person, not what the program did.
How can you tell whether an indicator is an output or an outcome?
Ask whether the number could rise while nobody's situation changed. If it could, it is an output; it is an outcome only when it describes a change in a person, household or organization you serve.
Whether the indicator is a count, a percentage or a quotation does not decide it. Partner A's “42 people placed in a job within 90 days” is a count and an outcome; “participants rated the sessions useful” is a percentage and, at most, a quality measure.
| Rule | Output | Outcome |
|---|---|---|
| Subject | The program: sessions, services | A person or organization served |
| What it records | Delivery or attendance | A change in state or behavior |
| Timing | During delivery | After, within a stated window |
| Unit | Sessions, people served | People who changed, per person |
| Missing answers | Rarely an issue | Reported as unknown |
| Can it rise with no change? | Yes | No |
Some indicators are neither. “Participants reporting a barrier to work” is context that helps explain an outcome; label it that way rather than forcing it into one column.
How do you word an outcome indicator?
Name the people, the change, the window and the source in one line, such as “completers who started paid work within 90 days of exit, from the placement survey”. Then write it as a dictionary row, so every partner counts it the same way.
| Vague outcome | Worded as an indicator | Denominator |
|---|---|---|
| Employability improved | Placed in a job within 90 days of exit | Completers |
| Stable employment | Same job at 12 months, reported annually | People placed |
| Better jobs | Hourly starting wage at placement, by track | People placed, per track |
| Lives changed | Name the change first; not yet an indicator | Not defined |
Measure the change at a point when it could have happened, against a baseline or comparison chosen before collection. Use a method that fits: a skills check, a service record, an observation or a well-written question, with open answers to explain the numbers.
A dictionary row turns wording into a counting rule: definition, dimension, optional standard code, roll-up, collection point, data type, disaggregation and source. The chapter A shared data dictionary: metric, dimension, standard shows how to write and test each row.
PROMPT · PASTE INTO CLAUDE, CHATGPT OR YOUR AI TOOL
Below is the list of indicators we report to [FUNDER]. 1. Label each one output, outcome or context, and say why in one line (could the number rise while nobody's situation changed?). 2. For each outcome, rewrite it in one line naming the people, the change, the window and the source, and give its denominator. 3. Flag any output worded to sound like an outcome. Rules: - Use only what our list says. Do not invent numbers or windows. - Where a window, source or denominator is missing, write "not in our data" and a question for us to answer. - People who did not answer a follow-up are unknown, never "no". [PASTE INDICATOR LIST]
What goes wrong when outputs are reported as outcomes?
The report looks full while the change goes unmeasured, and outcome numbers from different partners stop adding up. A funder reading the report against its agreement usually finds the gaps line by line.
In the fictional fund, Partner C counted enrollment as agreed: 80 unique people, a correct output. Its outcomes drifted. Placement was counted within six months instead of 90 days, retention at 12 months was missing, and starting wage came as one average instead of by track.
The fund sent three questions: “Can you count placements within 90 days, as agreed?”, “When will 12-month retention be available?” and “Can you split starting wage by track?” The outputs were never in doubt; every question was about an outcome.
| Mistake | What it looks like | Correction |
|---|---|---|
| Renaming an output | “Participants trained” becomes “lives changed” | Define the change in practice, then measure it |
| Skipping delivery quality | Completion reported as proof of good service | Add a quality or experience measure |
| Claiming cause from timing | Results rose after the program, so the program did it | Compare with earlier cohorts or outside data |
| Hiding the spread | One average across groups or tracks | Split by group; show who fell back |
| Moving definitions | 90 days this year, six months next | New definition, new row; keep a change log |
What can an outcome indicator prove, and what can't it?
It shows what changed among the people you measured; it does not show that the program caused the change. Hiring conditions, other services and who chose to enroll can all move placements. A causal claim needs a comparison planned before collection.
Follow-up is self-reported and incomplete. If some completers never answer the placement survey, you know how many answered yes, not the status of the rest, so report them as unknown and check whether they differ from those who replied.
AI can sort indicators and draft wording, but people decide what an outcome means and check each drafted number against the record. In Sopact Sense, every line of an AI Assistant answer links to a record you can open, and each person keeps one ID from the first form, so a 12-month follow-up lands on the same record as enrollment.
Start with one outcome for one program
Choose the outcome your funder asks about first, and give it the same care you already give your output counts.
- List every indicator you report and label each output, outcome or context with the six rules.
- Pick one outcome and reword it in one line: people, change, window, source.
- Write its denominator and its missing-value rule.
- Check that the follow-up form reaches people inside the window, on the same ID as enrollment.
- Report it next to the outputs that explain it, with the unknowns shown.
After the first cycle you have one outcome measured per person, the outputs that explain it and an honest count of who you did not hear from.
Frequently asked questions
What is an example of an output and an outcome?
In a job-training program, “participants who completed training” is an output: it counts what the program delivered. “Completers who started paid work within 90 days of exit” is an outcome: it describes a change in each person's situation, over a stated window. The output helps explain the outcome, since fewer completions usually mean fewer placements, but it cannot stand in for it.
Can an output also be an outcome?
It depends on whose work you are describing. A policy adopted can be an outcome of an advocacy campaign and an input to a later program that carries the policy out. State the scope of your results chain, then apply the test: could the number rise while nobody's situation changed? If yes, within that scope it is an output.
Are outcomes always long term?
No. Outcomes can be short, medium or long term. In the workforce example, placement within 90 days of exit is a short-term outcome, retention at 12 months a medium-term one, and living-wage jobs the long-term aim. Set each window to when the change could reasonably appear, and write it into the indicator.
Do outcomes have to be numbers?
No. People's own accounts can show an outcome and explain how it happened, such as what changed in their work after training. Collect them with a planned question at a set point, read them systematically, and report how many people answered. Keep themes counted among commenters separate from rates for everyone served.
Are outputs less important than outcomes?
No. Outputs show what was delivered, to whom and how well, and without them an outcome is hard to interpret. A drop in placements reads differently if completions also fell. Outputs fall short only when a report uses them to claim a change they cannot show, such as counting completions as proof of new jobs.
Does a higher outcome score prove impact?
No. A higher score shows change among the people measured. Before calling it impact, consider the quality of the measure, who answered and who did not, and other explanations such as local hiring conditions. A comparison, such as earlier cohorts or outside data on the same definition, is the minimum for a careful claim.
