Build a shared outcome matrix, combine only comparable results and make missing evidence visible. A practical Portfolio Intelligence course exercise.
By Sopact · Updated September 12, 2026
To measure outcomes across a portfolio, agree which changes matter, define comparable measures and keep each result linked to its population, period and evidence. Combine figures only when their definitions support it. Show different outcomes side by side when a single total would hide their meaning.
This practical lesson is for investment, grant and impact teams preparing a portfolio review. Bring the agreements from the onboarding lesson and the mapping register from the data dictionary lesson. Leave with an outcome matrix and a reporting rule for each measure. You do not need a new dashboard to complete the exercise.
“Show our impact” is too broad to guide collection. Choose a decision that the investment committee, board or delivery team needs to make. For example: which workforce partners need help improving sustained employment, and what evidence would justify that support?
That question changes what you request. Training attendance tells you about delivery. A follow-up observation tells you whether someone is working later. Their explanation may reveal a scheduling barrier, an unsuitable role or a change in circumstances. Keep the delivery count: it helps explain implementation and cost. Add the outcome evidence rather than deleting the operational data.
If you need a basic definition of outputs and outcomes, use the introductory guide. This chapter focuses on the next task: deciding what a portfolio may combine and what it should report separately.
Use one row per outcome and population, not one row per attractive metric label. Start with the outcomes stated in each partner’s agreement. Ask the partner whether those outcomes still reflect the work and what participants consider important.
| Portfolio question | Outcome to observe | Evidence to request | Boundary to retain |
|---|---|---|---|
| Are placements lasting? | Employment at an agreed follow-up point | Dated follow-up status and verification method | Eligible cohort, follow-up window, missing responses |
| Are earnings improving? | Change in earnings for the same people | Comparable baseline and follow-up observations | Currency, pay period, hours worked, matched records |
| Who is not benefiting? | Differences in outcomes between relevant groups | Appropriately collected group characteristics and outcome records | Consent, access, sample size and disclosure limits |
| What needs to change? | Barriers or unintended effects that affect progress | Participant feedback, partner explanations and supporting records | Who responded, when and whose voice is missing |
This is a fictional workforce example, not a required indicator set. A health, housing or environmental portfolio needs different outcomes and evidence. A common review format does not require everyone to measure employment.
A diverse portfolio often needs three levels: a shared question, measures appropriate to each group of similar organizations, and additional measures that matter to an individual partner. Forcing every partner into one outcome percentage can erase the purpose of the investment.
What change are we trying to understand?
Which partners measure the same outcome on the same terms?
What supports each observation and explains the difference?
For instance, partners offering job placement may share a job-start measure. A partner focused on retaining people already in work may need a retention measure instead. Both can contribute to a discussion about livelihoods, but starts and retention are not interchangeable counts.
Impact Frontiers’ five dimensions provide useful review questions: what changed, who experienced it, how much and for how long, what contribution was made, and what risks could affect the result. The framework also distinguishes investor contribution from enterprise contribution. Use those distinctions to avoid treating an investee’s observed result as proof of the investor’s causal impact.
Continue the three-partner example from the dictionary lesson. You have clarified their definitions and now request employment status six months after program completion. The following figures are fictional and do not describe Sopact customers.
| Partner | Eligible completers | Follow-up received | Employed at six months | Can it join this calculation? |
|---|---|---|---|---|
| A | 100 | 80 | 48 | Yes, subject to evidence checks and the agreed window. |
| B | 50 | 40 | 28 | Yes, if definition and method match A. |
| C | 60 | 45 | 30 | No: C measured at three months. Report separately until a six-month observation exists. |
First calculate coverage. A and B have responses for 120 of 150 eligible completers, or 80%. Their combined observed employment rate among respondents is 76 ÷ 120 = 63.3%. The 76 observed employed people are 50.7% of the full eligible cohort, but the employment status of the other 30 people is unknown. Neither figure is an estimate of what all nonrespondents experienced.
Do not average the percentages without checking denominators. A’s respondent rate is 60% and B’s is 70%. Their simple average is 65%; the combined respondent rate is 63.3%. A partner-weighted average may answer a different question, but it must be named and justified rather than presented as a person-level rate.
Keep C visible. Excluding its three-month result from this calculation does not mean excluding the partner from the report. Show it in a separate row with its observation window and plan for the later follow-up.
Before calling this a unique-person count: check whether anyone appears in both A and B. The example assumes no overlap. Resolve duplicate participation with an appropriate identifier and access rules; do not share identifiable records across partners simply to make counting easier.
A reporting rule makes the calculation repeatable when staff or reporting periods change. Save it in the dictionary, link it to the agreed collection questions and keep the prior version when a definition changes.
| Rule | What this example records |
|---|---|
| Outcome and unit | Employment status at six months; unique people and a clearly named rate |
| Population and timing | Eligible program completers; agreed six-month observation window |
| Numerator and denominator | Observed employed respondents / all respondents with usable status |
| Coverage | Respondents / eligible completers, reported separately |
| Exclusions | Three-month observations and incompatible definitions; reasons retained |
| Evidence and review | Source reference, observation date, collection method, reviewer and unresolved questions |
For qualitative evidence, retain the question, source and date alongside the analysis. A theme reported by ten respondents is not automatically the experience of the whole cohort. Read comments that contradict the dominant explanation as well as those that support it.
A later employment observation can show status at that point. A matched baseline and follow-up can show change in the observed group. Neither alone establishes that the program caused the change, or that the investment caused the program’s result.
In your review, separate three statements: what was observed, what might explain it, and what further evidence is needed. A partner’s account of an employer closing is relevant context; an AI-generated explanation is a hypothesis to check. Preserve both favorable and unfavorable observations, and record uncertainty rather than converting it into a confident score.
The next diligence lesson uses this distinction before investment: which claims have supporting evidence, which rely on assumptions, and what should be tested after funding?
Kuramo Foundation’s published story describes putting its gender-lens framework into a working dashboard across its accelerator, warehousing facility and fund. The practical lesson is to connect the framework to collection and review from the start. The story does not establish that every indicator is comparable across those platforms or that a dashboard proves causal impact.
Apply that lesson to your portfolio: decide which questions span the whole portfolio and which need separate views. A common system should preserve those differences while making the evidence easier to find.
Watch the portfolio reporting demonstration · 2 minutes 16 seconds
A Sopact demonstration of connected portfolio records and reporting. Use the outcome matrix in this lesson to decide what the portfolio report should show.
Close the exercise by preparing a short review note. State the question, compatible results, coverage, exclusions, explanations supported by evidence and the next action with an owner. Then ask each partner only for the clarification or follow-up that the decision needs.
When testing Sopact, use this complete sequence with your own sample records: a survey response, a supporting document and a later follow-up for the same partner. Ask whether you can trace a reported result and see unresolved evidence. AI can help review and summarize supplied material; people remain responsible for definitions, interpretation and approval.
They can share portfolio questions and review rules while using different outcome measures. Combine only compatible measures; report other outcomes in separate groups or views.
No. Outputs help explain delivery, coverage and cost. Pair them with evidence about changes for people or the environment rather than presenting activity counts as the outcome itself.
No. An unknown result is not a zero result. Report coverage, identify missing observations and explain how exclusions affect the interpretation.
AI can draft candidates from supplied plans and evidence. The portfolio team, funded organizations and relevant stakeholders must decide which outcomes matter and whether collection is appropriate and feasible.
Prepare an outcome matrix, a definition and aggregation rule for each shared measure, and one review note that shows coverage and uncertainty. Carry unresolved evidence questions into diligence and the reporting agreement.
Next: review outcome claims during pre-investment due diligence. Bring the evidence questions and reporting rules you have just written.
Start with one organization and a reporting question. Agree how measures compare, preserve source evidence, and make gaps visible before combining results.
Explore Impact & ESG Portfolio →
