Read the numbers and the explanation together. Build a source table, resolve apparent contradictions and write a review brief that makes uncertainty visible.
By Sopact · Updated September 12, 2026
Analyze an investee report by aligning the reporting period and definitions, then reading its narrative, measures, financial information and outcome evidence against the same questions. Record which sources support each conclusion, where they disagree and what remains unknown. Resolve apparent contradictions before combining results. End with a short review brief that separates observed findings, interpretation and the next action.
This lesson is for portfolio and impact teams reviewing recurring updates from investees or grantees. Bring the reporting package, dictionary and unresolved-gap register from the previous chapter. The exercise produces an evidence table and a review brief. It is an impact-reporting review, not a financial audit or an investment recommendation.
Start with the review question, not a request to summarize everything. “Did this partner deliver the agreed activity?” and “Did people sustain the intended outcome?” require different evidence. A report may answer one well and leave the other unresolved.
Use the outcomes agreed earlier in the course. Name the measure, observation window and decision the review supports. If the investment agreement calls for a six-month follow-up, a quarter’s attendance count cannot replace it.
Principle 6 of the Operating Principles for Impact Management connects monitoring with expected impact and an appropriate response when progress differs. The method here organizes that review; it does not turn a self-reported update into independent assurance.
The four perspectives are ways of asking questions. They are not four independent votes for a conclusion. A narrative may repeat a number from the same spreadsheet, and an outcome can be reported quantitatively or qualitatively. Trace each statement back to its underlying observation.
| Perspective | Question to ask | Common limitation |
|---|---|---|
| Narrative | What does the partner say changed, and why? | An explanation can be plausible without being established |
| Quantitative measures | What was counted, among whom and over what period? | A higher count may reflect a changed definition or population |
| Financial information | What resources were recorded, for which work and period? | Budget, expenditure and cash movement are not interchangeable |
| Outcome evidence | What changed for the intended people or environment, and how was it observed? | Activities and positive testimonials alone do not establish an outcome for everyone |
Do not add these perspectives into one score unless you have a justified scoring method. A complete document set can still provide weak outcome evidence. Conversely, a useful outcome observation may arrive with an administrative gap that needs clarification rather than dismissal.
Record each material claim alongside its source and scope. Capture the original statement before normalizing labels. Use page, sheet, row or transcript references that another reviewer can reopen.
| Claim or value | Source and scope | Review status | Question |
|---|---|---|---|
| Reported result | Exact location, organization and observation period | Supported, unclear or inconsistent with another source | What does the evidence actually establish? |
| Reported explanation | Who supplied it and what experience it covers | Partner interpretation or independently corroborated explanation | What would distinguish it from another explanation? |
| Financial figure | Document, currency, accounting period and category | Comparable with the requested measure or clarification needed | Is it the right basis for this comparison? |
| Unresolved gap | Expected observation and follow-up history | Unknown, unavailable or not yet due | What conclusion must remain limited? |
If a document is scanned or a table is extracted by AI, check important values against the original. A misplaced decimal, a lost negative sign or a shifted column can change the interpretation. Keep the source accessible to the reviewer rather than treating the extracted text as the only record.
Use the shared dictionary to identify units and definitions. Record incompatible measures separately. Mapping two labels to the same heading does not make their underlying observations comparable.
Fictional example, continuing the previous collection exercise. Partner A’s quarterly narrative reports 40 job starts and says employer outreach improved placements. Its spreadsheet contains 35 people who began work during the quarter and five accepted offers with start dates next quarter. The financial schedule records 120,000 in the reporting currency for the period. A three-month follow-up includes 20 respondents, of whom 14 report still being employed.
| Initial reading | What the evidence supports | What still needs checking |
|---|---|---|
| 40 job starts | 35 current-quarter starts plus five future offers in the supplied records | Partner confirmation and correction under the agreed definition |
| Outreach caused the improvement | The partner proposes outreach as an explanation | Comparison evidence and other possible influences |
| 120,000 spent this quarter | The schedule records that amount on its stated basis | Scope, currency and whether it covers the same activity and period |
| 70% retained employment | 14 ÷ 20 = 70% among the follow-up respondents | Eligible cohort, response coverage and how employment was verified |
You cannot assume the 20 respondents are the same people as the 35 new starters. A three-month follow-up may concern an earlier cohort. Confirm the contact identifiers and eligibility window before linking those observations.
You also cannot divide 120,000 by 35 and label the result “cost per sustained job.” The spending may cover other activities, and 35 is a start count rather than a sustained-employment count. First define the cost boundary and outcome window. A mathematically possible ratio is not necessarily a meaningful measure.
A defensible first conclusion is: “The submitted records distinguish 35 starts in the quarter from five future offers. The reported start count needs confirmation. Employment retention is 70% among 20 follow-up respondents, but cohort coverage is not yet established. The outreach explanation remains a hypothesis.”
Check scope before inferring error. Two figures can differ because they use different dates, populations, definitions or versions. Record both sources and ask a question that makes the difference visible.
For Partner A, ask: “The narrative reports 40 starts, but five spreadsheet records have next-quarter start dates. Does the agreed quarterly count exclude those five offers?” Retain the response and any restatement with the original evidence.
If the difference remains unresolved, report it as unresolved. Do not average conflicting figures merely to produce one number. A material concern should go to the designated reviewer under the organization’s process, with evidence and uncertainty stated together.
A useful brief tells the reader what can be concluded now, what cannot and what happens next. It does not hide important uncertainty in a footnote while presenting a confident headline.
For the fictional case, the action is to confirm the current-quarter count, establish the follow-up cohort and clarify the financial scope. It is not to declare that the program succeeded or failed. The next chapter compares accepted evidence with the impact agreement after these boundaries are clear.
AI can help extract candidate claims and organize references across documents. Give it the agreed question and definitions. Require it to identify missing information rather than fill gaps, and inspect important values and interpretations before using them in a report.
Practical review prompt:
Review the supplied reporting package for [organization] and [period] against [question and dictionary version]. Produce a table with claim, original value or quotation, source location, population, period, supporting or conflicting evidence, and unresolved question.
Keep observed results, forecasts and partner explanations separate. Do not infer missing values, causal effects or comparability. If sources disagree, show both and identify which definition or scope needs checking. End with findings supported by the supplied evidence and a list of human review actions. Do not make an investment recommendation.
Supply only information you are authorized to use in that system. A source reference in an AI response is a starting point for verification, not proof that the interpretation is correct.
Watch the portfolio reporting demonstration · 2 minutes 16 seconds
A Sopact demonstration of connected portfolio records and reporting. This demonstration introduces connected portfolio reporting. Use the source table in this lesson to test how evidence is preserved and reviewed.
Use one partner’s authorized reporting package and the source table above. Test whether metrics, files, narratives and follow-up observations remain attached to the right partner and period. Apply your dictionary and review rubric, then check whether a question about a discrepancy returns the relevant evidence.
The practical value is continuity: the next report can be reviewed alongside earlier observations and corrections. Confirm supported file types, import routes, access permissions and analysis configuration. Reviewers remain responsible for approving consequential conclusions.
You can summarize it, but that does not verify the underlying measures or explanations. Review the relevant source data and definitions for conclusions that depend on them.
Check period, population, unit, definition and version. Ask for clarification and preserve both sources. Do not choose or average values without a justified basis.
No. Several documents may repeat the same underlying observation. Identify the original source and assess what it establishes, rather than treating repetition as independent confirmation.
They can inform the same review, but ratios require matching scope and timing. Define which costs and outcomes belong together before calculating or comparing them.
No. Observed change and causal contribution are different questions. State the evidence for each and do not present a partner’s explanation as established causation.
Resolve material questions, retain accepted evidence and record remaining uncertainty. Then compare the results with the agreed expectations and decide the next action with the responsible team.
Next: compare investee results with the impact agreement. Bring the review brief and distinguish confirmed results from unresolved questions.
Bring one investee report, its supporting evidence and your review question. Work through the evidence and reporting workflow with Sopact.
Explore Impact & ESG Portfolio →
