Additional course links
This optional lesson is for people preparing an SROI analysis who already have a defined outcome and supporting evidence. You will build an adjustment register: one row per outcome, with the estimate, source, reasoning, uncertainty and reviewer. You can complete the exercise in a spreadsheet before configuring any software.
Start with the claim, population and period
Write the outcome precisely before estimating an adjustment. “Participants improved” is too broad. “Twenty participants reported sustained employment at the six-month follow-up” gives you a population and observation point, but still leaves questions about the definition of sustained employment, response coverage and other support received.
Keep the observed result separate from the explanation for it. An outcome record shows what was observed or reported. An adjustment is an analytical judgment about the claim you can make from that evidence. Preserve both so another reviewer can challenge the judgment without losing the original observation.
Keep three questions separate
| Adjustment | Question to investigate | A limitation to record |
|---|---|---|
| Deadweight | What change would likely have occurred without this activity? | A participant’s expectation is a perspective, not an observed counterfactual. |
| Attribution | Who or what else contributed to the outcome? | Naming another service does not establish its percentage contribution. |
| Displacement | Did this gain reduce an outcome elsewhere? | A participant may not know what happened to people outside the program. |
Social Value International’s guidance on avoiding overclaiming addresses the need to consider what would have changed anyway and the contribution of others. The practical task is to document the basis for these judgments, not to select reassuring labels.
Collect evidence that can answer the question
For deadweight, review whether you have a credible comparison, relevant external evidence or stakeholder accounts about likely alternatives. Check whether the population, period and circumstances are comparable. A published rate from a different setting may help frame a scenario without being a defensible estimate for your own participants.
For attribution, ask about other services, personal efforts and events that contributed. Retain the explanation alongside any requested estimate. If people cannot reasonably allocate percentages, do not make them invent precision. A qualitative account of contributions can be more useful than an unsupported number.
For displacement, identify who else might be affected and which evidence could reveal that effect. A question to participants alone may be insufficient. Record the boundary of the assessment: which people, organizations, places and time period were considered.
Across locations, agree the small common set of definitions needed for the analysis. Local teams can use different supporting questions. Map compatible evidence through a versioned dictionary; do not treat differently worded accounts as equivalent percentages without reviewing their meaning.
Build an adjustment register
Use a separate record for each outcome and model version. Include the outcome definition, eligible population, observation period, adjustment type, proposed value or range, source reference, method, limitations, reviewer and approval date. Distinguish a measured estimate from a stakeholder estimate and an explicit scenario assumption.
“Not assessed” is a valid status. It is not the same as zero. Zero asserts that no adjustment is needed; missing evidence says you do not yet know. If you use zero in a scenario, label it as a scenario assumption and show how the result changes under other plausible values.
Work through one illustrative calculation
Suppose a fictional model has $100,000 of gross outcome value for one year. For teaching only, assume 20% deadweight, then 25% attribution to others on the remainder, then 10% displacement on the adjusted remainder. These percentages are invented scenario inputs, not recommended defaults.
| Step | Calculation | Value remaining |
|---|---|---|
| Gross value | Starting amount | $100,000 |
| Deadweight | $100,000 × 0.80 | $80,000 |
| Attribution to others | $80,000 × 0.75 | $60,000 |
| Displacement | $60,000 × 0.90 | $54,000 |
Do not simply subtract 20 + 25 + 10 from 100. The example defines each adjustment against the remaining value. Your model must document its own bases and prevent overlapping adjustments from removing the same effect twice. Duration, drop-off and discounting are later assumptions; they are not included in this one-year example.
Test uncertainty before approving the result
Holding the other assumptions constant, 10% deadweight gives $60,750; 40% gives $40,500. The purpose of this range is to show sensitivity to one assumption. It is not a statistical confidence interval.
Identify which uncertain assumption changes the decision most. Then choose a proportionate response: gather better evidence, narrow the claim, report a range or defer monetization. A new survey question can improve the evidence available without automatically making the estimate reliable.
Keep analysis connected to its sources
In a Sopact workflow, plan the recurring collection and source records first. Keep participant or organization identity, observation dates and approved definitions connected where appropriate. Configure the analysis to retrieve relevant evidence and identify missing context. A reviewer still needs to judge whether the evidence supports a percentage.
The useful output is a reviewed register and a reproducible calculation. AI can help organize explanations or locate source passages; it should not silently convert narrative into causal estimates. Restrict access to identifiable accounts and use an appropriate summary for reporting audiences.
Practice: review one uncertain adjustment
- Choose one outcome and write its population and period.
- Separate the observed evidence from the proposed adjustment.
- Record what is known, what is assumed and what remains unassessed.
- Calculate at least two plausible scenarios using explicit bases.
- Ask a colleague whether the claim remains useful and defensible with those limitations visible.
Frequently asked questions
Does a participant survey prove attribution?
No. It provides evidence about the participant’s experience and view of contributions. Whether that supports a numerical adjustment depends on the question, context and other evidence.
Can an unassessed adjustment be recorded as zero?
Keep it marked as unassessed. If a calculation uses zero as an assumption, identify that assumption and test alternatives rather than presenting zero as a finding.
Must every course learner calculate SROI?
No. This is optional specialist work when monetary valuation serves a defined decision. An evidence-based outcome report may be the more appropriate output.