By Unmesh Sheth, Founder and CEO, Sopact. All fund, company and numerical examples are fictional.

Portfolio aggregation is a governed mapping decision. A company measure contributes to a portfolio measure only when its definition, population, unit, period and calculation support that use. Similar wording can suggest a candidate mapping; it cannot approve it.

Cedar's companies have different outcomes. Employment, stable housing and enterprise performance should remain distinct. Nina also wants a carefully defined service-reach view across the portfolio. That shared question requires its own mapping and overlap rules.

Define the portfolio question first

Write the portfolio measure before searching company dictionaries. Specify the eligible companies, qualifying service, period, unit, treatment of missing data and any deduplication boundary.

For this exercise, Cedar's reach measure accepts people completing a qualifying direct service during Q2, counted once within each company. It does not establish cross-company uniqueness. The output must therefore be labeled as summed organization-reported reach, with overlap unknown.

Build the mapping register

AI can search definitions for possible relationships and prepare a comparison. A reviewer approves, rejects or requests clarification. Save the source metric version, portfolio metric version, inclusion rule, rationale and approval date. Use a visible tag or linked mapping so a reviewer can identify which company measure feeds each chart.

A company may contribute to several thematic groups. That does not mean its value should be added repeatedly in a grand total. Keep multi-label classifications separate from mutually exclusive allocations. If allocating funding across overlapping themes, document weights and check that the allocation reconciles to the funding amount.

Calculate the result and its coverage together

The approved Q2 sum is 120 + 80 = 200 organization-reported people reached. Two of three expected companies contribute approved reach observations. MarketLink contributes no known value and is not assigned zero.

This result does not establish 200 unique portfolio beneficiaries, a portfolio-wide outcome rate or the effect caused by Cedar's funding. Two investments in RouteWorks do not double its 120. Without a defensible cross-company deduplication method, retain the overlap limitation.

Rates need numerator and denominator data. Do not average company percentages unless that is the intended statistic and the weighting is explained. A combined participant rate generally needs the compatible numerators summed and the compatible denominators summed, with overlap and coverage checked.

Cumulative reporting needs additional care. A sum of quarterly person counts may count the same person in several quarters. State whether the measure is period activity, quarter-end status or distinct people over a longer window.

Produce a calculation explanation

For the Q2 reach output, the explanation should identify:

• The question, reporting period and included companies.
• The two approved company observations and their source references.
• The definition and mapping versions used.
• The formula 120 + 80 = 200.
• The missing MarketLink result and the 2-of-3 company coverage.
• The unknown cross-company overlap and the decision not to claim uniqueness.
• The reviewer, approval date and report snapshot using this calculation.

This explanation is part of the output. A source link on its own does not tell a reader why one metric was included and another excluded.

Use the IRIS+ add-on when a report requires framework mapping

The existing data dictionary article and videos provide the detailed IRIS+ examples and cross-framework comparison. Keep that work attached to the dictionary version and report requirement. This core lesson concerns which approved observations can be combined.

Framework mapping is a separate comparison against an authoritative definition and its version. A thematic SDG association, an IRIS+ metric correspondence and a reporting obligation are different relationships.

Use working labels such as Exact, Related and Organization-specific, with Pending when evidence is insufficient. These are internal documentation labels, not official certifications. Record the reference, edition, scope differences and reviewer. Consult the GIIN IRIS+ catalog and the relevant framework's official materials. Do not invent a code or infer an exact match from its name.

If a donor requests an unfamiliar standard, obtain the applicable requirement before mapping. The course does not establish regulatory compliance or equivalence among frameworks. Keep the original company definition intact even when it supports more than one reporting view.

Build this part of your plan

Write Cedar's reach definition and mapping register. Produce the 200 calculation explanation. Then give RouteWorks a second thematic tag and show why the grand total remains 200. Describe what evidence would be needed for a unique-person total.

Check your reasoning: job starts and visits are not added to reach. MarketLink remains missing. A company can appear in several filtered views without being counted several times in the unfiltered total.

Test a tool: offer two similarly named metrics with different populations. Ask for the proposed mapping and explanation. The tool should preserve the mismatch for human review, and an unapproved mapping must not populate a released portfolio chart.

Next: Report from approved evidence.