The anatomy of an IC memo
that actually compounds.

The first IC memo I ever wrote was 38 pages, opened with a 4-page executive summary, and closed with a recommendation no one disagreed with. It was approved unanimously. Eighteen months later, on a different deal, no one could find it.

That has always been the strange thing about the IC memo. Every firm has one. Every analyst spends the worst three weekends of their year on one. And almost none of them compound. The next deal in the same sector starts from a blank document. The next associate doing diligence on the same target re-builds the cap table from the same data room. The institutional record that should be the firm's biggest asset reads, at the level of any single deal, like it was written for a single audience on a single afternoon.

The memo that doesn't survive

Most memos die at decision. They are written for the room. The room reads them on a Sunday evening, the room votes on a Tuesday, and the memo never opens again. The work survives in the analyst's head; the artifact does not survive at all.

What replaces it, six months later, is a folder. Inside the folder, the data room export, the deck, the model, and the memo, sitting next to each other like settled silt. The folder is searchable in the sense that you can find it. It is not searchable in any sense that helps the next deal.

"Every figure in the memo traces back to a document. Almost none of the figures trace forward — to the next memo, the next deal, the next thesis test." Field note · VectorShift Research

The compounding memo solves for the second half. It assumes its primary reader is not the IC; it is the analyst, twelve months from now, working a different deal in the same sector. It treats every claim as a small obligation to that future reader.

Three things a compounding memo does differently

We have looked at three years of IC artifacts across a dozen funds. The memos that survive — the ones that get re-opened, copied, and argued with — share three structural moves.

1. They name the thesis primitives. A compounding memo doesn't say "we believe in vertical SaaS for dental practices." It names the four or five primitives the thesis depends on — insurance-rail integration, single-instance hosting, the practice-management-system installed base, the regulatory tailwind from CMS — and says which two will be tested first.
2. They cite forward as well as back. The strong memos identify, in plain text, which numbers in the memo will only resolve in the first eighteen months of ownership. They name the report, the metric, and the threshold that will tell the firm whether the thesis is intact.
3. They mark the open joints. A compounding memo says "we do not yet know X, and we are buying anyway because the asymmetry is Y." Joints make a memo legible to the next deal — they are the place the next analyst's work starts.

The skeleton we recommend

The literal skeleton is less important than the disposition behind it, but for the firms we work with we recommend the following six-section structure. It maps cleanly onto the artifacts VectorShift inherits and surfaces in subsequent diligence.

Recommended sections

• Thesis primitives. Four to six numbered statements. Each one a single sentence. No qualifiers.
• The market frame. What is true about the sector that was not true five years ago. Two paragraphs at most.
• The asset. What you are buying and why this version of the asset is acquirable now.
• Returns architecture. Entry, the operating moves, exit. Not the model — the math behind the model's assumptions.
• What we don't know. The open joints. Named, ordered by what will resolve first.
• The thesis test cadence. When you will know — month 6, month 12, month 18 — whether the thesis is on track.

The skeleton is short enough to live as a template. Most of the firms running it have a two-line version of each section as a default in the memo system, with the analyst's draft replacing each line.

What VectorShift inherits from the memo

This is where the artifact starts paying compounding interest. The thesis primitives — once named and stored as discrete claims — become first-class objects the platform can track. When a future deal touches the same primitive, the memo surfaces automatically. When a portfolio company misses one of the named thresholds, the deviation alert points to the IC paragraph that named it.

The crude version of this is grep. The honest version is a structured representation:

// IC memo — Acme Dental, May 2026
thesis_primitives: [
  { id: "insurance-rail-integration", test_by: "month-12", threshold: "≥3 carriers live" },
  { id: "single-instance-hosting", test_by: "close+90d", threshold: "100% migrated" },
  { id: "pms-installed-base", test_by: "month-18", threshold: "≥18% share" }
],
open_joints: [
  "competitive response from Dentrix on insurance-rail product"
]

The shape is not the point — your firm's representation will look different. What matters is that the memo's claims become queryable artifacts. A year later, when an analyst on a different deal asks "how have our dental memos handled the insurance-rail thesis," the answer is a list, not a search.

Where the thesis stays open

The hardest part of writing a compounding memo is not the structure. It is the discipline of leaving the thesis open at the right joints. Analysts are trained to close arguments. The firm needs the analyst to close the ones that should be closed, and to flag — clearly, in a section the IC will read — the ones that should not be.

The asymmetry is real: a memo that names its open joints reads, on the day of the vote, less convincing than one that doesn't. The bet is that on day 451 it reads more useful than any other artifact the firm has on the asset. We think the bet is worth taking.