What is portfolio data management?
Portfolio data management is the practice of collecting, defining and combining the evidence funded partners report, so every portfolio figure can be traced to the partner, period, definition and record behind it. It covers partner surveys, reports, documents and the rules for adding results.
This guide is for funders, networks and intermediaries who collect results from many partner organizations and need one picture they can defend. (The term also describes investment-market data, which is not covered here.)
THE SHORT VERSION
- Give each partner its own workspace, where it builds its own surveys and governs its own data, while the funder sees the combined results.
- Agree a small shared core of definitions before the first submission, and let every other question stay local to the partner that asks it.
- Roll up counts from records, never percentages or partner-calculated totals, and show missing and non-comparable partners beside every figure.
Why does partner reporting break at quarter end?
Partner reporting breaks because each partner collects in its own template, and the data meets the funder’s definitions only at the end, when someone reconciles it by hand. A program officer spends a week matching columns and asking which quarter a number covers.
The usual fixes move the problem: a CRM or grants portal needs a central model built by an admin who eventually leaves, and a warehouse needs data engineers most funders do not have. The result is six systems and zero trust, and an AI assistant on top is only as good as the mess underneath.
Take a fictional funder, the Northfield Skills Fund, which supports four partner organizations running skills training. Each reports one core measure, people completing the agreed training program, and each asks local questions: Partner B about shift patterns, Partner D about childcare.
How do folders let each partner govern its own data?
Give each partner organization its own folder, where its team builds its own surveys and asks the AI Assistant about its own data, while the funder, as owner, sees aggregated results across every folder. In Sopact Sense, folders work as team workspaces.
At Northfield, Partners A to D each get a folder. Partner B adds its shift-pattern question without waiting for the funder or an administrator. Its assistant sees only Partner B’s participants; the fund’s portfolio lead sees completions across all four.
Write down who creates folders and who approves access, so the rules outlast any one staff member; lesson 4 of the free course, Govern the work, turns them into a roles checklist. Field selection keeps names, emails and phone numbers away from AI models in every folder.
Which definitions should every partner share?
Share only the few fields you will combine, define each one once with its unit, population, time basis and evidence rule, and let every other question stay local.
SHARED CORE · NORTHFIELD EXAMPLE (ILLUSTRATIVE)
Core measure: people completing the agreed training program; enrollment does not count.
Unit and population: unique eligible participants within this award.
Time basis: each completion recorded with its date, never a partner’s own cumulative total.
Missing value: not reported, not applicable or unavailable, with a reason. Blank is never zero.
Definition change: a dated entry in the team’s change log, approved by the definition owner.
Two partners can use the same words and still count differently: one counts a participant who missed the final session, another does not. Partner B’s shift question and Partner D’s childcare question stay in their own folders.
If a framework requires particular indicators, map to them and record the differences; the IRIS+ catalog can inform a dictionary, but a shared label does not make numbers addable. Context management, a shared place for definitions the AI Assistant can use, is coming soon in Sopact Sense; until then, one named person keeps the dictionary as a team document.
What changes when you govern partner data at collection?
Governing at collection means the shared core is built into each partner’s forms from the first submission, so the funder’s roll-up counts records instead of reconciling spreadsheets. Agreeing on meaning happens once, at the start.
Four design principles carry it. Collect flexibly: surveys, reports and attachments sit in the partner’s folder. Keep every wave: each participant gets one ID at the partner’s intake form, and completion and follow-up attach to that record with no merge.
Analyze on arrival: an Intelligence Cell reads each narrative report as it lands, with a prompt your team configures, and returns the barriers it names with the passage attached. Govern every step: partners run their folders, and the Assistant stays locked until someone declares which surveys it may use.
Public data follows the same rules: Northfield loads county employment benchmarks as an ordinary survey, so partner results sit beside local conditions without a separate pipeline. Keep two dates on every submission: when it arrived and the period it covers.
How do you roll partner results into one portfolio figure?
Add counts under the shared definition, never percentages or partner-calculated cumulative totals, and put coverage, missing partners and non-comparable measures beside the figure.
NORTHFIELD, YEAR ONE · FICTIONAL
Under the old templates, Partner A reports 30 completions cumulatively at quarter one and 50 at quarter two: year to date is 50, not 80. Partner B reports 12 and 18 as period-only counts: year to date is 30.
Partner C has not submitted quarter two. Partner D reports enrollment, not completion.
The honest portfolio sentence reads: “Partners A and B report 80 completions year to date (50 + 30); Partner C is missing for quarter two, and Partner D’s enrollment is shown separately.” It is a sum of partner-reported completions, not necessarily 80 unique people.
Then Partner A corrects quarter one from 30 to 28 and confirms the cumulative 50, so the quarter-two increase becomes 22, not 20. Record the correction, its reason and its approver in the change log.
In year two, with the shared core built into each folder, every completion is a dated record on a participant’s ID. The fund counts records for any period, so the cumulative question disappears and Partner C’s gap shows as missing coverage.
Some checks remain: flag a rate entered as 180 percent, keep both accounts of a disagreeing report, and ask the partner. Because a participant can enroll with two partners, decide whether you count people or completions; the deep dive Many programs, one picture works through overlap and coverage.
Some limits no workflow removes. Partner results are self-reported until someone checks the evidence, a missing partner stays unknown, and an AI-proposed theme stands only when a named person accepts it. A rise across the portfolio does not show your funding caused it; that claim needs a comparison designed in advance.
How do you evaluate portfolio data management software?
Judge each option by who governs partner data after launch and whether a portfolio figure can be rebuilt from records, not by its form builder or dashboard. Each setup does its own job well; the difference is who can change it.
On a narrow screen, scroll the table horizontally to read all columns.
| Setup | Who governs it | Can a partner add a local question? | What the roll-up rests on |
|---|---|---|---|
| Emailed templates and a spreadsheet | Whoever holds the file | Yes, in its own copy | A manual merge every quarter |
| CRM or grants portal | An admin or consultant | Usually through the admin | The central model and its exports |
| Data warehouse with BI | Data engineers | Through a pipeline change | Pipelines that need upkeep |
| Governed collection with partner folders | Each partner team; funder as owner | Yes, in its own folder | Records collected under the shared core |
A grants portal or CRM remains the right home for applications, awards, payments and relationships. The question here is where partner results land and who can change a form.
Test each shortlisted option with one figure you have already reported, including a corrected value, a missing partner and a hard-to-read document, and have a second reviewer rebuild it from the records. Ask one partner’s assistant about another partner’s participants; a good setup cannot answer.
Ask the same question twice and expect the same count, then open one line of the answer to check it reaches a record. The portfolio monitoring buying guide extends this test into ongoing monitoring.
Where does Sopact Sense fit in a portfolio?
Sopact Sense is where partner evidence is collected and governed: a folder for each partner, one ID from the first form, analysis on arrival and answers whose every line links to a record.
That also covers field selection, declared scope, public data loaded as an ordinary survey, Intelligence Cells and Rows, and an Assistant that answers with sources; Claude or ChatGPT can query the same records through MCP. It does not replace a grants payment, accounting or CRM system; keep those and use the portfolio workflow for the evidence around them.
Open Play Foundation runs four sports facilities in Stellenbosch, South Africa, covering coaching, water infrastructure and food security. With its data governed at collection, ten program reports became one funder submission, and a water leak surfaced in real time. CEO Marco Botha: “I’m digitizing our entire business through Sopact.”
Watch governed collection in a working product
This introduction shows Sopact Sense keeping every participant and portfolio connected from application to exit, with analysis on arrival and answers that cite their evidence. Watch how a new response attaches to its record and analysis starts there, not from an export.
Start with one workflow this week
Pick one shared measure and two or three varied partners, and run one full reporting cycle before you scale anything. The real test is the second cycle, after a correction arrives and a partner adds a question.
- Write the portfolio question and the decision it serves, such as: “How many people completed the agreed program this year, by partner, and what got in the way for the others?”
- Give each pilot partner a folder, and name who creates folders and who approves access.
- Write the shared core: three to five fields, each with its unit, time basis, evidence rule and owner. Mark every other question as local.
- Build the intake form where each participant gets an ID and the completion form that attaches to it; let partners add their own questions.
- List the fields that must never reach an AI model, at least name, email and phone, and choose the surveys the Assistant may use.
- At cycle end, build the roll-up from counts with coverage beside it, and have a second reviewer rebuild it from the records.
After the first cycle you should have a folder per partner, a written shared core with an owner, a change log, and one portfolio figure whose every line opens a record.
Frequently asked questions
What does portfolio data management include?
For a grant or impact portfolio, it includes partner organizations and agreements, reporting periods, shared measures and their definitions, partner surveys, narrative reports and documents, quality checks, corrections, access rules and the rules for combining results. The aim is that any portfolio figure opens down to the partner, period and record behind it, with gaps visible.
Is portfolio data management the same as portfolio monitoring?
No. Data management keeps the records and their meaning in order: who reported what, under which definition, for which period. Monitoring uses those records to review progress against plans during the year, and analytics looks for patterns across comparable records. Both are only as reliable as the data management underneath them.
Can each partner see only its own data?
In Sopact Sense, yes, when each partner works in its own folder. The partner’s team builds its surveys there, and its AI Assistant sees only that folder’s data, so it cannot answer questions about another partner’s participants. The organization owner, usually the funder or network lead, sees aggregated results across every folder.
Can we keep existing partner templates?
Often, for a transition year, if their fields and periods map reliably to your shared core. Test a sample first: templates tend to hide cumulative totals, blanks that mean different things and enrollment counted as completion. Moving the core into forms each partner fills in its own folder removes the quarterly mapping, and partners keep their local questions.
Can quarterly cumulative results be added together?
Usually not. A cumulative quarter-two figure already includes quarter one, so adding them counts the same completions twice. Use the latest cumulative total or the change between periods, and check that the definition and population did not shift. Recording each completion as a dated record lets you count any period directly.
Does AI remove the need for reconciliation?
No. AI can read partner narratives on arrival, pull out reported figures and flag apparent inconsistencies, but it cannot make two definitions equivalent or confirm that a claim is true. Reconciliation shrinks when definitions are agreed before collection and answers trace to records, and a named person still reviews results before they reach a funder report.
