Tech-powered portfolio management for venture.

Every asset class has an operating model that suits it. Public equity has prices, so the hard problem is judgement. Private credit has contracts and schedules, so the hard problem is servicing. Venture has neither a price nor a schedule, and its portfolio problem is therefore a data problem: knowing, at any moment, what you own in each company and how each company is doing.

Most venture firms start out managing that with spreadsheets, and for a first fund that is a reasonable choice rather than a lazy one. The argument here is not that spreadsheets are bad. It is that three structural features of venture — none of which a manager can change — make manual portfolio management degrade in a particular way: slowly, invisibly, and worst at exactly the moment the information matters most.

Three features that make venture different

1. Breadth

A venture portfolio holds many small positions rather than a few large ones. That is not an accident of style; it is the shape the return distribution demands. The operational consequence is that per-company attention is the scarcest resource in the firm. A partner cannot sit inside twenty or forty companies the way a buyout deal team sits inside three. Attention has to be directed, and directing it requires a comparable view across the whole book — which is the one thing manual tracking is worst at producing.

2. Illiquidity and the subjectivity of the mark

There is no price feed. A private company's value between financing events is a judgement, and the judgement has to be evidenced: a methodology, the inputs used, the date, the person who approved it, and the reason it differs from the last one. Nothing about that is controversial in principle. In practice, when marks live in a spreadsheet, the evidence lives in email threads and in memory. The number survives; the reasoning does not. A year later nobody can reconstruct why a company was held flat through two quarters and then written down.

3. Heterogeneous reporting

Early-stage companies have no finance function. A seed company's monthly numbers are produced by a founder in the evening; a Series C company's are produced by a controller. Both are in your portfolio and both report to you, and their data quality is not comparable. Add to that the ordinary variation in definitions — revenue recognised differently, headcount including contractors or not, runway computed off last month's burn or an average — and a portfolio-level total becomes a sum of things that are not the same quantity.

Why the failure is quiet

If manual portfolio management failed loudly, it would already have been fixed. Nothing breaks. No cheque bounces. No system goes down. The failure mode is simply that the firm learns things later than it could have, and the cost of learning late never appears anywhere as a cost.

This is the counterfactual problem. A reserve deployed a quarter after it was needed still shows up as a reserve deployed. A company written down at the next round rather than when the deterioration began still gets written down; the mark simply moves in one step instead of three. A pro-rata right allowed to lapse leaves no trace in any report — there is no line item for the position you did not take. The portfolio you have is visible. The portfolio you would have had with better information is not, and it is never the subject of a post-mortem.

Where it actually goes wrong: the ownership ledger

If there is one place to look first, it is the record of what the fund owns. Not the headline percentage from the last round, but the ledger underneath it — position by position, instrument by instrument, event by event. This is where manual tracking most often turns out to be wrong, and the reason is structural: venture ownership changes through a long series of events, each of which is individually simple and collectively unforgiving.

• Multiple priced rounds. Each one dilutes existing holders on terms that depend on the pre-money valuation, the size of the round and any secondary component alongside it.
• SAFEs and convertible instruments. These sit off the equity ledger until they convert, and then convert on terms — caps, discounts, whether the conversion is pre- or post-money — that materially change everyone else's percentage. A cap table that ignores unconverted instruments overstates the fund's ownership, sometimes considerably.
• Option pool expansions. A pool created or refreshed as part of a round dilutes existing holders, and whether it is taken pre- or post-money decides who pays for it.
• Pro-rata decisions. Exercising, partially exercising or waiving a right changes the position and the ownership path from that point on. The decision needs to be recorded as a decision, with its reasoning, not inferred later from the resulting numbers.
• Secondary transfers and structured holdings. A position bought or sold outside a financing round, or held through a vehicle alongside the main fund, has to reconcile to both the company's register and the fund's own books.

Each of these is easy to record once. The difficulty is that they compound: an instrument mis-modelled at conversion silently corrupts every ownership figure and every return calculation that follows it, across every report, until somebody reconciles to the company's register and finds it. The place this is discovered is usually a transaction — a follow-on, an exit or a secondary — which is the most expensive moment available to discover it.

What the LP is actually underwriting

A venture LP cannot verify judgement. They can read your memos and your track record, but the picks themselves are unfalsifiable until the fund is nearly over. What an LP can assess, today, is your process — and so that is what much of diligence is really about. Where does the mark come from. Who approved it. How do you know a company is off plan. What happens between the miss and your response. How quickly can you answer a question you were not expecting.

These questions are not administrative box-ticking. They are the only observable evidence an LP has that the next fund will be run as well as the last one was described. A manager who answers them from a live record is making a different impression from one who answers them in a week, having rebuilt the numbers first. Portfolio monitoring that runs continuously is therefore a fundraising asset as much as an operational one, and the funds that treat it that way tend to say so in their operating history rather than in their deck.

What a fund should be able to answer on demand

A practical test, with no advance notice and nobody preparing anything:

1. What do we own in every company today, fully diluted, including unconverted instruments?
2. What is each position marked at, on what methodology, approved by whom and when?
3. Which companies are behind plan on revenue or burn this month, and by how much?
4. Which companies have under a defined number of months of runway remaining?
5. What is our remaining reserve, and against which companies is it earmarked?
6. Which pro-rata rights are live in the next two quarters?
7. What did each company report last period, and does it agree with what we told our investors?

Every one of those is answerable from a spreadsheet, given time. The test is not whether the answer exists. It is whether it arrives in minutes without a person assembling it — because that is the difference between a portfolio you monitor and a portfolio you manage.

The argument, briefly

Venture does not need technology because technology is modern. It needs it because the asset class combines wide portfolios, subjective marks and uneven reporting, and those three things make a manual record fall behind reality without announcing that it has. Systems do not improve judgement. They make sure that when judgement is exercised, it is exercised against a current and defensible picture of what the fund actually owns.

That is a modest claim. It is also the whole difference between finding out in month two and finding out in month five.