Organize partner reports, documents and metrics across reporting cycles—with clear records, shared definitions, quality checks and a worked reconciliation example.
Portfolio data management is the process of organizing, checking and maintaining the information used to understand a portfolio. For foundations and impact funds, that means keeping each funded organization’s agreements, reported results, documents and review history usable across reporting cycles.
The term also appears in investment-market data and enterprise data-asset management. This guide focuses on partner and impact evidence: grantee reports, investee metrics, financial attachments, interviews and the definitions needed for a reliable portfolio view.
A good process lets an authorized reviewer move from a portfolio figure to the organizations, periods, sources and calculations behind it. It also makes missing or incompatible data visible. That requires more than a shared folder, but it does not require every record to live in a single software product.
A partner can participate in two programs, receive several awards and submit multiple reports. A report can contain many measures, and one measure can change definition over time. Treating all of this as one row creates ambiguity about what a figure describes.
Define the records and their relationships before importing a large history. A persistent organization ID helps connect a renamed partner to its earlier reports. Separate agreement and reporting-period references prevent two awards or two quarters from being mistaken for each other.
Illustrative identifiers. One organization may have several agreements; each agreement may have many observations.
An ID does not prove two records belong together. Establish matching rules, review ambiguous matches and keep a crosswalk between source-system IDs. Do not merge organizations merely because their names look similar, or treat a person’s changed email address as a new organization.
The important distinction is maintained structure versus repeated reconstruction. A well-managed spreadsheet can preserve identifiers and history. A poorly configured database can still duplicate records. Judge the process by whether it keeps the relationships intact.
A data dictionary records what a field means and how to interpret it. For each shared measure, agree the unit, population, reporting window, calculation, source and owner. Include examples of what belongs in the field and what does not.
On a narrow screen, scroll the table horizontally to read all columns.
| Field | Illustrative definition | Why it matters |
|---|---|---|
| Measure | People completing the agreed training program. | Enrollment and attendance are not substituted for completion. |
| Unit and population | Unique eligible participants within this award. | Multiple sessions do not create multiple people. |
| Time basis | Cumulative from award start to the quarter end. | Quarterly cumulative figures must not be summed. |
| Evidence | Accepted completion record under the agreed rule. | A submitted claim can be distinguished from a reviewed result. |
| Missing value | Not reported, not applicable or unavailable, with a reason. | Blank is not automatically zero. |
| Version | Definition v2 effective from an agreed date. | Changed rules do not silently rewrite earlier periods. |
Keep partner-specific measures where they are relevant. Only map them to a shared portfolio field when their meanings are compatible. If a funder requires a particular framework, document the mapping and any differences; a familiar indicator label is not enough.
The IRIS+ catalog provides impact metrics that can inform a dictionary. Your team still needs the applicable reporting context and agreed collection rules. Continue with the portfolio data dictionary and standards-mapping lesson for the course treatment.
Inventory the data before choosing an import. Useful portfolio evidence may arrive through partner surveys, quarterly templates, diligence files, financial statements, social audits, site notes and exported interview or Zoom transcripts. Some material sits in a CRM; some is stored in Google Drive, Dropbox, Box or SharePoint.
For each source, identify the owner, access permission, update frequency, format and authoritative record. Decide whether the workflow uses a submitted file, a scheduled export or a configured integration. Verify connector availability and behavior rather than assuming a named file store is already connected.
A transcript may help explain a reported change, but its date, participants and permissions matter. A financial statement can support financial information; it does not by itself establish a social outcome. Connect sources for interpretation without pretending they are interchangeable evidence.
Clear questions, required fields and allowed formats prevent some errors at the source. They do not eliminate incorrect reporting, changed definitions or contradictory documents. Use validation to identify issues, then a review process to resolve them.
On a narrow screen, scroll the table horizontally to read all columns.
| Check | Example | Handling |
|---|---|---|
| Completeness | A required reporting period has no submission. | Show it as missing and assign follow-up. |
| Validity | A percentage is entered as 180. | Check scale and definition before accepting it. |
| Consistency | The narrative and attachment report different totals. | Preserve both and request clarification. |
| Uniqueness | The same submission is imported twice. | Detect the duplicate using source and record references. |
| Timeliness | The latest file describes an old period. | Retain it, but do not label it current evidence. |
| Comparability | One partner counts temporary roles, another sustained employment. | Keep separate or establish a justified common measure. |
Record who resolved an exception, what changed and why. A plausibility flag is not proof of an error: a large change may be real. Conversely, a value within an allowed range may still be wrong. Important claims need appropriate source review.
Track the quality of the process as well as the results: missing submissions, recurring definition questions, unresolved conflicts and corrections after publication. These indicators help you improve collection without blaming partners for unclear requests.
Illustrative example. Partner A reports 30 completions cumulatively at the end of quarter one and 50 cumulatively at the end of quarter two. Its year-to-date result is 50, not 80. The increase during quarter two is 20, provided the cohort, definition and accepted historical values are consistent.
Partner B reports 12 completions in quarter one and 18 in quarter two, explicitly as period-only counts. Its year-to-date result is 30 if those counts can be added without duplicate completions. Partner C has not submitted quarter two.
Partners A and B report 80 completions year to date: 50 + 30. Partner C is missing from this result. This is a sum of partner-reported completions, not necessarily 80 unique people across the portfolio.
Before calling that a unique-person total, check whether a person can complete programs run by both partners and whether the collection design supports an appropriate deduplication method. If it does not, retain the honest unit and explain possible overlap.
Now suppose Partner A corrects quarter one from 30 to 28 while confirming that quarter two’s cumulative 50 is unchanged. The implied quarter-two increase becomes 22. Keep the correction, its approval and the reason; do not overwrite history and leave readers unable to explain why the change calculation moved.
These checks are needed whether the data sits in a spreadsheet, CRM or specialist platform. Stable references reduce matching work, but they do not remove the need to reconcile meaning.
Give a named data owner responsibility for the dictionary, source mapping and acceptance rules. Give program staff a clear way to correct records, request changes and resolve routine exceptions. Reserve structural or sensitive changes for the appropriate review.
Define access by role and purpose. A partner may need to submit its own evidence, a program officer to review assigned awards, and a portfolio lead to inspect approved aggregates. Access to an aggregate should not automatically expose confidential attachments or individual-level information.
Operational independence means staff can perform authorized tasks reliably. It does not mean bypassing security, privacy or finance controls. The aim is a manageable set of rules that survives a staff handoff.
Choose a portfolio figure your team has already reported and try to reproduce it from source evidence. Include inconsistent templates, one corrected value, a missing partner and a document that is difficult to read. Use the same sample for each shortlisted setup.
A plain-language assistant may phrase an answer differently on a second run. The underlying approved calculation, selected records and definition version should remain inspectable. Do not substitute identical wording for reproducibility.
Ask which capabilities are available now, which need configuration and which require paid services. Include migration, connectors, document processing, training and ongoing review in the cost estimate. The portfolio monitoring buying guide extends this test into operational monitoring and implementation planning.
Configured AI analysis can help organize incoming text, identify candidate measures, extract reported figures and flag apparent inconsistencies. It is most useful when the result stays attached to the source, reporting period and agreed definition, ready for review.
Sopact’s portfolio workflow connects an agreed partner plan with recurring collection and source-linked reporting. Evaluate that workflow using your own partner survey, uploaded document and historical result. Confirm the supported formats, permissions and integrations in the proposed implementation.
Sopact is not a substitute for every CRM, accounting system or data warehouse. A foundation may keep those systems and improve the collection and interpretation of partner evidence around them. The relevant test is whether the complete workflow reduces reconstruction while making the evidence easier to inspect.
The video below is a broader portfolio-reporting companion that includes financial-holdings examples. Use it to explore multiple-source reporting; the partner-data model, quality checks and limitations are in this guide.
Kuramo Foundation’s story describes making an internal gender-lens framework operational through data strategy, collection and a dashboard before launch. It illustrates the value of connecting a framework to the work. It does not establish that reconciliation disappeared or prove a particular investment outcome.
For your own portfolio, start with three varied partners and one shared question. Write down which definitions differ, which sources are authoritative and which results cannot yet be combined. Then build the collection and review process around those findings.
Continue through the Portfolio Intelligence course, beginning with its foundation lesson and following the linked onboarding and dictionary lessons. For the reporting output, use the impact report writing guide and report examples and dashboards.
For an impact or grant portfolio, it includes organization and agreement records, reporting periods, measures, source documents, validation, corrections, permissions and the rules used to combine results.
No. Data management maintains the records and their meaning. Monitoring uses those records to review progress against plans and identify follow-up. Analytics examines patterns across compatible records.
No. IDs help match entities and submissions, but the data model, population definitions and overlap rules still matter. One organization may hold several agreements, and the same person may appear in more than one partner program.
Often, if their fields and reporting periods can be mapped reliably and the import format is supported. Test a representative sample. Missing or ambiguous evidence may still require a targeted request or a revised template.
Retain the original submission, accepted correction, reason, reviewer and date. Show which reports or calculations changed and which definition version applies.
Usually not. A cumulative quarter-two figure already includes the earlier period. Check the time basis and use the appropriate latest total or period change, with corrections and population boundaries accounted for.
No. It can help identify inconsistencies and prepare evidence for review, but it cannot make incompatible definitions equivalent or verify every submitted claim. Keep human review and source references for important results.
Choose one shared question and a few varied partners. Trace the answer through agreements, periods, definitions and sources, then address the gaps before scaling the workflow.
