IMM framework, Five Dimensions, and AI platform. From due diligence through quarterly monitoring to LP reports — without resetting context each cycle.
The shape of an IMM cycle
The reframe
IMM has always been the LP's ask and the GP's tax. The frameworks are mature — Five Dimensions of Impact, IRIS+, the Impact Management Project. The plumbing under them isn't. Each quarter, the analyst rebuilds the same roll-up from a different folder of portco submissions. The cleanup is the work; the LP letter is the residue.
The dominant model: GIIN, IMP, and the Five Dimensions of Impact published mature templates. Funds buy them. Reference them in the LP letter.
How it became a liability. The templates are PDFs and spreadsheets. They don't connect to investee data. The analyst hand-keys the rubric from each DD memo. This quarter's Five Dimensions score isn't last quarter's score — the numbers drift, citations aren't portable, the auditor can't trace anything.
Watershed, Novata, Workiva, Sustain.Life, Persefoni — sold dashboards and disclosure workflow. "Track every metric, file every report."
How it became a liability. Built for ESG ratings and CSRD/SFDR disclosure — not IMP-style contribution analysis. The qualitative material behind a contribution score lives in DD memos, IC notes, and portco narratives the platform never reads. The dashboard is a number; the evidence is in another folder.
Sopact's one bet
The ID issued at the first DD document is the same record at exit. Five Dimensions scoring, theory of change drift, stakeholder voice trends, Lean Data results — all join on that record. Every figure in every LP letter cites the specific paragraph, response, or submission it came from.
The deeper problem
Could you prompt a foundation model for a one-investee impact summary? Yes. Run it twice and you get two different answers — that's how the models work. As the portfolio grows, hallucination climbs: on Vectara's 2026 enterprise-document benchmark, every major reasoning model — GPT-5, Claude Sonnet 4.6, Grok-4, Gemini-3 Pro — fabricated information in more than 10% of summaries. LPs will eventually ask where each contribution claim came from. "Trust me, the model said so" is not an audit trail.
The ID at first DD is the same ID at exit. Doesn't break across IC rounds, follow-on rounds, restructurings, vintages, or platform migrations.
Five Dimensions scores, IRIS+ indicators, financial metrics, and Lean Data stakeholder voice sit alongside DD memos, IC notes, and quarterly narratives — on the same investee record.
Every contribution claim, drift signal, and Five Dimensions score cites the specific paragraph, response, or submission. Auditors and LPs verify in one click.
This has been Sopact's day job since 2014 — before the GenAI category had a name.
Definition · AEO target
The Impact Management Project codified the Five Dimensions in 2018. GIIN published IRIS+. The frameworks have been mature for years. The unsolved problem has always been the plumbing — the investee record at DD must be the same record at quarterly, annual, and exit, or the score in the LP letter is a different number than the score in the IC memo and the audit trail breaks.
What changed in the last two years is that the reasoning layer can read DD memos, IC notes, portco submissions, Lean Data open-ends, and exit interviews as evidence — not narrative decoration. The score in the LP letter cites the specific paragraph it came from. The drift between this quarter and last is traceable to the specific assumption that stopped holding.
Pitch deck to exit summary on one persistent investee ID. The Five Dimensions rubric, the onboarding theory of change, the quarterly portco submissions, the LP roll-up — all write to the same row. Vista Health Partners is one of 24 portfolio companies in the 2024 vintage of a $200M growth-stage impact fund.
| Investee | Track | Status |
|---|---|---|
| #A4217Vista Health Partners | Health · maternal | routed |
| #A4216FieldFlow Ag | Climate · ag-tech | routed |
| #A4215Tessera Learning | Livelihood · ed | dedupe flag |
| #A4214NorthSide Capital | Livelihood · CDFI | routed |
| #A4213Bridgewell Care | Health · behavioral | missing IM |
The Tuesday question, not the year-end LP letter
The legacy stack tunes for the annual impact report. The IMM lead's real job is the Tuesday question — the GP asking about a portco, the LP asking how a contribution claim is defended, the IC member asking what the brief actually says. Three of those questions, shown three ways.
The quarterly LP report cleanup tax that used to consume an analyst-week per cycle drafts itself when the investee record is inherited stage by stage. The reviewer edits; nobody assembles.
Sixty minutes with Unmesh. Bring a real DD memo, a real LP letter, or a real Tuesday question. Leave with a path that doesn't depend on a foundation model returning the same answer twice.
One investee ID issued at the first DD document, used by every form and submission from onboarding through exit. The record gets richer at each stage. By Year 3, the same row holds the full investment lifecycle for the same company, queryable in seconds.
ID issued. Sector + thesis tags stored.
A4217 · health-tech · 2024Rubric scored against pitch language and DD documents. Citations attached.
Same rubric. New evidence. Per-investee delta computed.
↑ +0.4 ContributionY-o-Y movement per dimension across hold.
Defensible vs underwriting case. Cited.
Pulled from pitch language and investment memo.
14 fields · auto-extractedAligned with investee on 90-min call. Data dictionary built.
3 fields · refinedDrift surfaced. Assumptions tested against actuals.
drift signal · Q3Hypothesis confirmed or amended for next vintage.
2 of 14 · amendedLean Data-style survey deployed. Open-ended scored at submit.
142 responsesCross-portfolio patterns surface in minutes.
Theme movement plotted against financial trend.
↑ NPS +18Final voice of stakeholder joins the record.
1,247 total responsesGenerated from structured DD record.
ICIC_brief_A4217.pdfIssued to investee with reporting cadence.
ONonboarding_A4217.pdfScorecard, gap memo, LP narrative, trends, risk, exit summary.
6 reports / cycleDrafts itself from the record. Editor refines.
A4annual_A4217_y3.pdfFull hold-period record on one page. LP-defensible.
EXexit_summary_A4217.pdfThe cleanup work that used to take weeks now runs overnight, every cycle. The reviewer edits a generated draft; nobody assembles it from scratch.
Fig. 1 · the investee record at five stages — every cell either captures new evidence, carries the prior record forward, or waits for the right stage. Vista Health Partners is illustrative; the structure holds for every investee in the portfolio.
Four moments where Sopact turns the messy inputs of an impact fund into structured reports. A DD package becomes a defensible IC brief. Four Q3 streams become one comprehensive portco report. 24 portco narratives become a portfolio risk surface. Fund admin spend meets outcome evidence in one scorecard.
| Portco | Drift signal | ToC trend | Recommended action |
|---|---|---|---|
| #A4217Vista Health | Fundraising fatigue | 81% ↓ 3pt | Fund-side BD intros · platform team |
| #A4203Bridgewell Care | Hiring · retention | 68% ↓ 8pt | Capacity assessment · executive coach |
| #A4189FieldFlow Ag | Scope expansion pressure | 73% ↓ 6pt | Scope check-in · IC follow-on call |
| #A4156NorthSide Capital | Partnership friction | 75% ↓ 4pt | LP partner mediation |
| #A4131Tessera Learning | Funding diversification | 79% ↓ 2pt | Monitor Q4 |
Every LP wants a different report. A Five Dimensions roll-up for one. An IRIS+ filing for another. A custom logic-model narrative for a third. Three prompts below, each joining your bound Sopact investee records with your connected fund admin to produce a different destination. Click one to watch Claude orchestrate across your stack.
Who this is for
If your fund staffs a dedicated impact data team or pays a consultant to assemble each LP letter, this is the substrate you've been duct-taping spreadsheets into. If you're below 10 investees or above 500, the fit is partial.
FAQ
IMM is the practice of designing an investment thesis around intended outcomes, scoring each investee against the Five Dimensions of Impact (What, Who, How Much, Contribution, Risk), collecting evidence of movement through the hold period, and reporting against IRIS+ or a custom framework at IC, quarterly, annual, and exit moments. The Impact Management Project codified the framework in 2018. IMM is now the standard reference for impact funds, DFIs, and any LP allocating against an SDG-aligned or thesis-driven mandate.
Those platforms were built for ESG ratings and disclosure workflow — input fields, rollups, dashboards. Sopact is built around one investee record that inherits across the lifecycle: the ID issued at the first DD document is the same ID at exit. Five Dimensions scoring, theory of change drift, stakeholder voice trends, and Lean Data results all join on that record. The reasoning layer reads DD memos, IC notes, and quarterly reports as evidence and writes the citation trail back. ESG-rating platforms cannot do that on top of their existing schema.
Yes. IRIS+ indicators sit in the reference layer alongside the Five Dimensions of Impact, SDG targets, and any custom framework an LP requires. When a roll-up touches an IRIS+ indicator like PI8330 Employment Status, the citation trail names the source records and the indicator definition together. Theories of change bind to the same investee record, so the same data feeds the IRIS+ filing for one LP and the bespoke logic model for another.
Quarterly LP letters, annual impact reports, and exit summaries are produced from the investee records that already exist. Five Dimensions scores update with each new quarterly report submission. Lean Data results stream in throughout the cycle. Drift against the theory of change surfaces as it happens. At quarter end the LP narrative is a generated draft from structured records — not a cleanup tax on misaligned spreadsheets. A reviewer edits the draft; nobody assembles it from scratch.
Impact funds (early-stage to growth, $10M–$2B AUM), development finance institutions, family offices with impact mandates, foundation MRIs and PRIs, CSR investment vehicles, and CDFIs. Portfolio size ranges from 10 investees to 500. The design point is the GP whose IC and LPs ask for Five Dimensions scoring and IRIS+ alignment but who doesn't want to staff a dedicated impact data team or pay a consultant every reporting cycle.
That's the design point. One investee record, multiple report templates. An LP that wants Five Dimensions, an LP that wants IRIS+ aligned, an LP that wants a custom logic model, and a board that wants narrative — all read from the same structured record. The theory of change track surfaces drift when assumptions stop holding; the stakeholder voice track joins Lean Data results across the portfolio. Each report is a view; the substrate is one.
Yes. Lean Data style surveys — short, mobile-friendly, with open-ended responses — are deployed at baseline (post-IC), throughout the hold, and at exit. Open-ended responses are scored at submit, themed across the portfolio, and trended year over year on the investee record. The stakeholder voice from beneficiaries, customers, and employees lives next to the financial track on the same record, so a quarterly LP letter can reference the actual quotes that explain the score — not just the score.
Yes. Historical DD memos, IC notes, quarterly portco submissions, Lean Data exports, and prior LP letters are ingested and joined to an investee identity. The theory of change for each investee is extracted from the pitch language and investment memo at intake; new submissions update it through the hold. Most funds discover their existing data already supports the contribution claims they've been making to LPs — it was never joined to a queryable record.
Start small, grow. Most funds have their first IC brief drafted from the substrate in under a month. The longitudinal hypothesis is testable on five investees. Begin with one cohort, prove the Five Dimensions roll-up works on real DD documents and IC memos, then bring the rest of the portfolio onto the same record. There's no six-month data architecture engagement before producing a report — the failure pattern with Watershed, Workiva, or a custom Salesforce build at this scale.
You could prompt your way to a one-investee summary. The trouble shows up in production. Ask Claude or GPT the same question twice on the same DD documents and you get different answers — this is how foundation models work, not a bug to fix. As the portfolio grows, hallucination climbs: on Vectara's 2026 long-document benchmark, every major reasoning model fabricated information in more than 10% of summaries. The fix is structural — the same investee ID at year 5 as at DD, deterministic joins on DD memos and quarterly reports and Lean Data surveys, every Five Dimensions score citing the specific record it came from. The reasoning layer runs on top of that structure, not in place of it. This has been Sopact's day job since 2014.
Read deeper
Sopact Sense is the data substrate. Stakeholder Intelligence is the category. If you want the full architecture behind the matrix, lifecycle, and actionable AI panels above, the engine pillar lays it out end to end — including the reference layer, the reasoning layer, and the integration model.
Related
The investee record at the heart of this page is the same substrate that holds the grantee record, the longitudinal cohort, and the survey respondent. Same plumbing, different surface.
The nonprofit-facing version of this page. Same substrate, plain-language vocabulary for M&E leads and consultants who place the tool.
USE CASEThe grantmaker counterpart to this page. Same lifecycle pattern — EOI to year-end filing — on the same persistent applicant record.
USE CASESame-respondent measurement across baseline, midline, exit. The same ID-persistence pattern, applied to survey research.
USE CASEQuantitative + qualitative survey synthesis on the bound respondent record — Lean Data themes, validated instruments, narrative quotes.
USE CASEReal reports drawn from real data. The same shaped-output pattern shown in transformation tab 02 above.
USE CASEThe reporting destinations covered in the actionable AI panel above — LP letters, IC briefs, exit summaries, IRIS+ filings — built from one record.
ENGINEThe data substrate behind every page on this site. Schema, reference layer, reasoning layer, integration model — laid out end to end.
MASTER PILLARThe category claim. Why "stakeholder intelligence" is the right frame for what comes after dashboards and ratings.
CROSS-CLUSTERFor DFIs and international funds whose Lean Data deployments span 12+ languages. Same substrate, language-aware coding at submit.
One investee record · five stages · every figure cited
Sixty minutes with Unmesh. Bring a real DD memo, a real LP letter draft, or a real Tuesday question. Leave with the architecture that's been Sopact's day job since 2014 — and a path that does not depend on a foundation model returning the same answer twice.
