By Sopact · Updated September 12, 2026

Portfolio onboarding turns a funding decision into an agreed reporting plan. The funder and investee or grantee decide which outcomes matter, what evidence is available, who will report it, and when it will be reviewed. Keep that agreement linked to later submissions so each quarter builds on the same record.

This lesson is for portfolio managers, investment teams and grant managers setting up reporting after approval. Bring the funding agreement, relevant diligence, an existing theory of change if available, and the evidence map from your course overview. Leave with a versioned reporting agreement and a tested first submission. The exercise supplies a fictional partner-reporting example; it does not require a particular software product.

1. Prepare from existing evidence.
2. Draft an evidence map.
3. Agree useful measures.
4. Specify necessary documents.
5. Approve a version and schedule.
6. Test one submission.
7. Resolve exceptions.
8. Report with boundaries visible.

What belongs in an impact agreement?

Here, an impact agreement means the measurement and reporting commitments agreed with a funded organization. It can be a reporting schedule or annex linked to the funding documents. Do not assume it replaces the signed investment or grant agreement. Resolve any conflict with the people responsible for that agreement.

The record should answer practical questions: what are we trying to learn, which observations can answer it, who owns the work, and what happens when a definition changes? A framework name alone cannot answer those questions.

1. Prepare from the evidence you already have

Start after approval, at a time agreed with the funded organization. A focused meeting may be enough; a complex investment may need several conversations. There is no universal 90-minute requirement.

Read the application, diligence memo, baseline, site notes, financial information and existing framework before asking the partner to repeat them. Separate commitments already agreed from proposals made during diligence. An analyst’s suggestion is not automatically a reporting obligation.

• List the outcomes and reporting requirements stated in the funding documents.
• Mark important definitions or evidence that remain unclear.
• Identify information the partner already collects and what a new request would cost.
• Agree how notes or a transcript will be captured, used and shared.

If you use an AI assistant to review documents, provide the actual authorized files or text. A pasted URL does not guarantee that the assistant can access its contents. Check the available file types and access permissions in your implementation.

2. Draft an evidence map, not a finished theory

A transcript or existing framework is a starting point. Extract the stated outcome, proposed indicator, source passage and unresolved question. Keep inferred links separate. Do not turn an aspiration such as “stronger livelihoods” into a precise target without discussion.

For example, a narrative may describe new jobs but omit hours, duration or the reporting period. That is a definition gap. A report may supply a number without supporting records. That is an evidence gap. Treating both as a red score hides what needs to be resolved.

Optional prompt: draft the questions for review

Use only the supplied documents for [organization].
Create a table with:
- stated outcome or commitment;
- proposed measure and exact source passage;
- definition, population and reporting period, where stated;
- evidence already available;
- unresolved question and suggested review owner.
Separate source statements from proposals. Write “not stated” where information is missing. Do not invent targets, classify outcomes as verified, or send messages.

Review the output against the source. The prompt helps prepare a conversation; it does not approve the framework or validate the evidence.

3. Agree useful measures with the partner

Discuss what matters to the people experiencing the outcomes and to the reporting audience. Keep outputs, outcomes and evidence of contribution distinct. Training attendance may be important delivery information, but it does not establish improved income.

Impact Frontiers’ Five Dimensions can help structure the discussion of outcomes, affected people, scale and duration, contribution and risk. Choose measures that fit the work. A shared framework does not require every investee to report an identical set of numbers.

For each proposed measure, agree the definition and the collection burden. Retain partner-specific measures where they are meaningful. Identify separately the small set of compatible measures needed for a portfolio view. The linked lesson develops the data dictionary and standards mapping.

4. Specify the documents you actually need

A quarterly update can combine quantitative metrics, narrative explanations and attachments. It is more than a survey score. Ask for financial reports, audit evidence or other documents when they are relevant to the agreement and decision—not because every partner must supply the same three files.

A social audit is not a universal requirement for grants or investments. Specify what the term means in this relationship, who prepares it, which period it covers and whether the agreement requires independent assurance. An uploaded document should remain a source to review, not a check mark that automatically proves compliance.

5. Approve a version and set the reporting schedule

Record the agreed version, approval date and effective period. “Locking” an agreement should mean protecting its history, not making it impossible to improve. Keep amendments visible and explain whether a revised definition applies prospectively or requires restating earlier figures.

Define the observation date separately from the submission deadline. A June 30 workforce count submitted on July 15 is still a June observation. Financial statements and outcome follow-ups may need different schedules. Do not force an annual outcome into a quarterly change claim just because the financial report is quarterly.

Set reminders and escalation rules only where the team has agreed them and the system supports the configuration. An AI-generated reminder draft is not a scheduled workflow. Test the recipient, timing and approval route before relying on it.

6. Test one submission before scaling the process

Ask one partner to submit a sample through the intended process. Include a metric, a short explanation and one supporting file. Confirm that all three remain attached to the correct organization, period and agreement version.

Test an incomplete value and a conflicting value as well as a clean submission. Check whether the review identifies the exact issue without silently replacing evidence. Where AI analysis is configured, inspect the source passages and compare its findings with a human review. A summary or score should not substitute for checking important claims.

1. Open the result and its supporting document from the same organization record.
2. Confirm the observation period and submission date are distinguishable.
3. Correct a value and check that the earlier value and reason remain available.
4. Test access with the submitter and reviewer roles.
5. Ask a colleague to reproduce one reported result without your explanation.

7. Resolve exceptions without rewriting the original

Compare the new submission with its agreement and prior period. Separate missing evidence, changed definitions and performance variance. A result below target may need explanation or action; it is not necessarily a data error.

Send a focused request that names the measure, period, problem and useful response. Agree a reasonable response date and use the escalation route when necessary. Do not assume weekly reminders are suitable for every organization or that more reminders fix an unclear definition.

We are reviewing [measure] for [period].
Your submission reports [value], while [source/definition] states [difference].
Could you confirm [specific question] or provide [specific evidence]?
If the requested information is unavailable, please tell us what is available and why.
Please respond by [agreed date] or contact [owner] to discuss timing.

Review the draft before sending. Preserve the original submission and record the clarification or correction separately.

8. Prepare a report with its boundaries visible

Only combine compatible observations. Show the reporting period, participating organizations, missing evidence and review status alongside the result. An unresolved report does not always prevent sharing an interim update; it does require a clear limitation.

Prepare the audience’s view from the approved evidence, then review access before distributing it. A board pack, LP report and partner feedback note may need different detail. Do not expose sensitive source material merely because a shareable link is convenient.

For structure, use How to Write an Impact Report and browse report examples. These support presentation, not verification of the underlying evidence.

Practice: finish the agreement for organizations B and C

Self-contained fictional exercise. The board needs women in paid full-time roles at June 30. B reported 60 temporary placements across six months without a gender breakdown. C reported 25 relevant roles, but for March 31.

• For B, ask whether the required June observation exists. Do not relabel the 60 placements or infer a gender breakdown.
• For C, request the June observation or mark it unavailable. Preserve the March value as prior-period evidence.
• For each organization, agree the future observation date, submitter, source, deadline and review owner.
• If collecting the requested measure is not feasible, discuss another measure or report the limitation. Do not invent historical data.

Deliverable: one approved reporting-agreement page per organization, plus an unresolved-questions log. A colleague should be able to identify the next due submission and explain why the original values cannot yet form a comparable total. Bring those definitions into the linked lesson.

Frequently asked questions

What is an impact agreement in portfolio management?

It is the agreed measurement and reporting plan for a funded relationship: outcomes, definitions, evidence, responsibilities, schedule and change controls. It should align with the funding documents rather than silently replace them.

Must every investee use the same reporting cadence?

No. Agree a schedule that fits each measure, the available evidence and the reporting obligation. Observation dates, submission deadlines and review checkpoints should all be explicit.

Does onboarding require a social audit?

Only where the relationship’s requirements and purpose call for one. Define its scope, method, period and review status. Do not treat the same document list as mandatory for every investment or grant.

Can AI approve the impact framework or send reminders automatically?

AI can help prepare drafts and identify questions. People approve the agreement and review important evidence. Scheduled reminders require a supported, configured workflow and an agreed sending or review policy.

What should happen when an agreed metric changes?

Record the revised definition, approval, effective period and reason. Explain whether historical observations remain comparable or need restatement. Preserve previous versions.

What is the next step after portfolio onboarding?

Build the data dictionary from the approved agreements. Record each measure’s definition and version, then decide which partner-specific measures can support shared portfolio questions.

Optional implementation reference: Build a portfolio data dictionary and map the relevant standards.