Insights · Research

Private markets software has a decision integrity problem

The gap sits between having data and knowing whether last quarter's investment decision is still defensible. Existing tools handle the first. Almost none of them continuously test the second.

11 min readUpdated October 2026Capital Refinery research

The short answer

Decision integrity is the property of knowing whether a prior investment decision still holds, given everything that has happened since it was approved. Private markets software today is organized around collecting data: market comps, CRM workflow, diligence documents, portfolio KPIs and fund accounting. None of those six layers holds the investment committee's thesis as a structured record it can keep testing. The thesis is written once, filed, and only reopened by hand when someone asks whether it still works.

$3.6Tunrealized value on GP balance sheets (Bain, 2025)
6layers in the dominant private markets software stack today
41document families in Capital Refinery's classification taxonomy

The premise

Each private markets investment starts as a structured argument. The deal team builds a thesis, identifies the conditions it depends on, and quantifies what has to be true for the entry price to make sense. The investment committee approves it, and the wire goes out.

What happens next is where the industry has a structural gap. The thesis becomes a slide deck on a shared drive. The conditions become a folder someone might reopen in year three. The reasoning behind them becomes one principal's memory, and that principal is usually staffed on a new deal by the time it matters.

Meanwhile the operator sends in updated KPIs each quarter. The accounting feed fires. The credit agreement terms get tested. None of it is connected to the original thesis in any structural way, so by the time anyone notices the thesis is broken, the position has already moved.

A $13 trillion industry has spent fifteen years optimizing the data layer. The result is a stack that knows everything about today and very little about whether yesterday's decision still works.

The six-layer stack

The dominant private markets software stack has six layers today. Each one is mature, well funded and strong at its job. None of them holds the investment decision as a structured object the platform can keep testing against new operator data.

01

Market data & screening

Sourcing, comparable transactions and fund benchmarks. The outside view of a deal, not the inside view of the position the firm actually closed.

PitchBookS&P Capital IQPreqinCoStar

02

Deal CRM & relationship

Pipeline, relationships and process workflow. The IC memo is an attachment here, not a data model.

DealCloudAffinityAltviaDealpath

03

Diligence / VDR

Hosting, securing and increasingly summarizing the diligence room. Mature inside the window; disconnected the moment the deal closes.

DatasiteIntralinksHebbiaBlueFlame

04

Generative AI for private equity

Chat over documents, useful for research and diligence acceleration. Not built to maintain a structured record of the decision.

HebbiaBlueFlameAlphaSenseEndexEilla

05

Portfolio monitoring

KPI ingestion, dashboard rendering and LP reporting for the post-close window. iLEVEL alone is reported in use at 700+ firms.

iLEVELAllvueChronographCobalt LP

06

Accounting & operations

The books, the waterfall and the LP statements. Mature, audited and regulated, and entirely below the decision layer.

FundCounteFrontStandard MetricsQBO

07

Decision integrity

The missing layer. Holds the IC thesis and conditions as a record the platform keeps testing.

Capital Refinery

The missing seventh layer: decision integrity

A decision integrity layer is not a UX skin over an existing tool. It requires its own extraction model, its own data model, and its own feedback loop back to the committee.

01

Deterministic-first extraction

Figures are traced to the source document where a locator resolves. Where candidates conflict, a local model chooses among them; it does not generate values from scratch.

02

The thesis recorded as a structured object

Memo sections are assembled from that sourced evidence at approval, not written once as a Word document and filed.

03

Conditions carried past the vote

Open conditions stay on the deal record alongside the covenant obligations register, instead of a line in an approval email.

04

Certificates tested against covenants

Each compliance certificate after close is tested against the covenant it evidences, with headroom and drift surfaced next to the KPI.

What changes when the layer exists

None of this replaces the six layers above it. It is what they were not built to be.

  • The IC memo is not only a slide deck headed for a shared drive; its sections are assembled from sourced evidence and stay attached to it.
  • Monitoring stops being only a visibility exercise: each compliance certificate and registered deliverable is tested against its covenant, with headroom and drift surfaced next to the KPI.
  • Open conditions stay on the deal record that monitoring reads instead of living in an email thread.
  • The memo exports as DOCX or XLSX carrying a deterministic evidence fingerprint, so it can go to an LP without being rebuilt from scratch.
IC memo screen
Live product screen · illustrative data

Why this layer is buildable now

Three shifts made this layer practical. The first is small, capable local language models that can run as a thin adjudication step over deterministic extraction, on the same server as the parsers. The second is the industry's growing acceptance that rebuilding the analysis from scratch each quarter is the wrong way to spend analyst time. The third is LP scrutiny: as exit windows extend, the question of whether a prior decision still holds has moved from a nice-to-know to something an LP will ask directly.

Capital Refinery is built around the investment decision as the unit of analysis: statements spread with figures traced to source where a locator resolves, ten advisory reconciliation checks run on arrival, memo sections assembled from that sourced evidence, and open conditions kept on the record monitoring reads after close.

Continue reading

  • Your data room dies at close. Your risk doesn't.
  • Why your monitoring dashboard isn't a decision tool.
  • Private credit just got its first real stress test.
  • The IC materials problem: rebuilt from scratch each quarter.
  • Time-to-consequence: the metric your portfolio tools don't have.
  • What is an IC memo, and why does it go stale?
  1. Bain and Company, Global Private Equity Report 2025: $3.6 trillion unrealized value on GP balance sheets.

Questions

What is decision integrity in private markets investing?

Decision integrity is whether an investment committee's original thesis and the conditions it depended on still hold, given everything that has happened to the position since approval. It is a property of the decision, not a KPI.

Why don't portfolio monitoring platforms solve the decision integrity problem?

Platforms such as iLEVEL, Allvue and Chronograph are built around the operator KPI as the primary object. None of them has a structured representation of the IC thesis to test the KPI against.

How many layers are in the private markets software stack?

Six today: market data, deal CRM, diligence and virtual data rooms, generative AI, portfolio monitoring, and accounting and fund operations. Decision integrity is the layer missing from all of them.

Does generative AI close the decision integrity gap?

Generative AI tools accelerate drafting and document review, but they are not built to maintain a structured, continuously tested record of the investment decision. Faster output is not the same as a structural fix.

How does Capital Refinery address decision integrity?

It spreads statements with figures traced to source where a locator resolves, runs ten advisory reconciliation checks on arrival, assembles memo sections from that sourced evidence, and keeps open conditions on the record monitoring reads after close.

Accelerate your deal execution process.

Private markets, connected. Bring automation to your deal execution.

Founder-led access. Every working session runs on your actual documents, not a generic demo.