Private Markets Don't Fit Systems Built for Public Securities

A multi-asset portfolio management system works beautifully until it encounters a PDF. Institutional funds spend tens of millions on high-speed networks to capture public market pricing in milliseconds. Their private equity valuations arrive forty-five days late as an email attachment. When the investment board raises the target allocation for private credit and infrastructure, the operations team does not check the system limits. They check a spreadsheet tracking unfunded commitments.

Public markets systems are built on a universal equation: quantity multiplied by price equals value. Data arrives daily via standardized feeds to update an investment book of record. Private markets do not have a daily price or a fixed quantity. They have commitments, drawdowns, distributions, and unfunded balances. Forcing a private equity fund into a public securities schema usually means creating a dummy ticker symbol, setting the quantity to one, and manually updating the price to match the total net asset value whenever a statement arrives.

A system built to multiply closing prices by share quantities will always choke on a recallable capital distribution. The core accounting engine understands a distribution as a reduction of cost basis. It does not understand that a portion of that distribution might be called back next quarter. To track the recallable unfunded commitment, the fund accountant builds a shadow spreadsheet, knowing the core system’s unfunded balance is now permanently wrong.

The mismatch goes deeper than the schema. A daily price is an observation. A quarterly mark is an opinion, produced by a process, delivered late, and subject to revision. Systems built for public markets treat the quarterly mark as a slow version of a daily price. They are not the same thing.

You cannot compute a daily risk number from a quarterly mark and pretend otherwise. When public equities drop ten percent in a week, the public data feeds into the risk engine immediately. The private equity valuations will not update for another sixty days. The risk engine suddenly shows the portfolio is massively overweight in illiquid assets. The denominator effect is a data frequency problem wearing a costume. You can only disclose the gap.

The PDF is not the problem. The PDF is a symptom of a data model that does not fit the asset, combined with a power dynamic that prevents it from being fixed. In public markets, exchanges and regulators force standardized reporting. In private markets, the general partner holds the leverage. They have no commercial incentive to change their bespoke reporting formats to make the limited partner’s IT operations easier. The highly subscribed funds simply refuse to use standardized templates. The limited partner accepts the PDF because the alternative is losing the allocation.

Enterprise software vendors sell unified multi-asset dashboards. When you look under the hood of how the private asset data gets into that dashboard, you find an army of operations staff manually transcribing numbers from general partner PDFs into staging tables. The dashboard is modern. The data pipeline is from 1998. Vendors sell the dashboard, but they bill you for the manual data entry hidden behind the API.

The system of record for private markets at most institutions is the spreadsheet that one person maintains and three people know about. It is not the vendor platform, and it is not the custodian’s portal. The platform is the reporting layer.

The spreadsheet exists because it holds decisions and workarounds the core platform has no fields for. Every side letter is a data field that lives in a legal PDF, containing most favored nation elections, fee breaks, excuse rights, and co-invest rights. If a side letter is not in the data model, it is not in the report, and it is not in the fee validation. It is in a folder.

The spreadsheet also reconciles the three clocks that govern private markets: the date an economic event occurred, the date the source reported it, and the date the institution recorded it. A capital call notice arrives with instructions, dates, and amounts that several teams must interpret differently. Treasury needs to make the payment. The investment team needs to update the remaining commitment. Accounting needs to determine what portion is an investment, a fee, or an expense. They are looking at the same notice but need different entries and different controls. The spreadsheet holds the mapping between them.

Optical character recognition and extraction tools compress the ingestion step. They get the number off the page. They do not get the number into the books. Extraction does not resolve reference data, reconcile cash to the bank, handle restatements, or apply side letter terms.

When an extraction tool works perfectly in a pilot using historic PDFs, it will fail on day one of the new quarter when a major infrastructure fund changes its branding and shifts its data tables half an inch to the right. The tool fails quietly, mapping carried interest figures into the return of capital field. Automation can remove rekeying without removing judgment. Someone still has to validate the document version, map the manager’s terminology, resolve exceptions, and decide how a corrected notice changes prior records.

Fixing this is not a software procurement project. It is building a permanent factory. The useful cost question is not what it takes to buy a private-markets module. It is what it takes to run the module when every manager has different notices, definitions, corrections, and delivery schedules.

Implementing a mid-size institutional platform takes nine to twenty-four months. The first close cycle after go-live will be worse than the spreadsheet. The second will be comparable. The third will be better. Plan for three.

The ongoing cost is mostly headcount, not software. You have to budget the run function. Someone has to own the ingestion, the exceptions, the general partner chasing, and the reconciliation. The limited partner’s books and the general partner’s books will never reconcile perfectly. The tolerance policy is a governance decision, not a technical one. The process is not to eliminate the differences but to explain them within an agreed tolerance.

Getting data faster or in a better format is a negotiation, not a configuration. It happens one general partner at a time, over years, and depends on the limited partner’s leverage. The custodian does not eliminate this work. Some custodians will take private markets data as a service, but the process is often the same manual ingestion with a different logo on the invoice. The custodian’s advantage is the control environment and audit trail, not automation.

The platform becomes the system of record when the investment team stops keeping its own spreadsheet, which takes two close cycles and one audit.