Somewhere in the past few years, close speed became the thing private capital finance teams tell each other about. Fourteen days down to nine. Nine down to five. A few platforms now talk credibly about a three day close, and they are not exaggerating.
It is a real achievement by people who worked hard for it, and we have no interest in diminishing it. We do think it is a transitional advantage. Firms treating close speed as their finance strategy are building on ground that will have levelled out well inside a single fund cycle.
The reason is simply that speed converges. Every constraint that made the close slow is being attacked at once, by administrators, by the systems their clients run, and now by models capable of reading the source documents that previously required a qualified person and a fortnight. When the whole sector arrives at the same answer, the answer stops carrying information about you.
Every input to a fast close is becoming available to everybody
Consider what actually makes a close slow. Bank data arrives in inconsistent formats. Capital activity sits in one system and the general ledger in another. Someone has to work out whether a particular expense is chargeable to the fund or the management company, and the answer lives in a limited partnership agreement written six years ago and modified by four side letters. Valuations arrive late from portfolio companies whose own finance functions are thinner than yours. Intercompany balances disagree by amounts nobody can explain quickly.
None of that is a mystery, and none of it is proprietary. Bank feeds standardise. Administrators invest in the same integrations for every client they have, because that is how their margins work. Document extraction against agreements and side letters is now reliable enough that the reading of them is no longer the rate-limiting step. Within a few years, a five day close will be unremarkable for any competently run mid-market platform, and firms still taking twenty days will not be showing operational sophistication so much as an unaddressed backlog.
The advantage in this sort of convergence is real but temporary. It accrues to whoever gets there first, and it expires when the second and third arrivals show up. Nobody builds a durable franchise on an advantage with a known expiry date.
A fast number and a defensible number are not the same achievement
What we see in practice is that acceleration and explanation come apart. A close gets faster because exceptions get resolved faster, and exceptions get resolved faster because someone experienced makes a judgement and moves on. The judgement is usually right. It is also usually undocumented beyond a note in a reconciliation file and whatever that person remembers in six weeks' time.
That gap surfaces later, and always at the least convenient moment. An operational due diligence team asks how a particular management fee offset was calculated. An LP's finance function, running its own extraction across every GP it backs, notices that the treatment of transaction fees appears to have changed between the third and fourth quarters, and asks why. The auditors want the basis for a valuation input that was agreed verbally in a Tuesday call. In each case the number is right, and in each case retrieving the reasoning behind it takes days of somebody senior reconstructing what they were thinking.
A firm that closes in three days and then spends eleven days a quarter answering questions about the close has not built the advantage it thinks it has. It has moved the work downstream to people who are more expensive and less available.
The durable asset is the explained close
The version worth building is a close where every figure arrives with its own evidence attached. Not a reconciliation that ties, which is the current standard, but a reconciliation that states in plain language what the number is, where each component came from, which document or clause governs its treatment, and what judgement was applied where the source material was ambiguous.
In practical terms, a traceable figure carries four things:
the source document or system record, cited to the page or the transaction
the governing clause where a legal term determines the treatment
the judgement applied, stated as a sentence a non-accountant can read
the person who accepted that judgement, and when
This is where the choice of technology stops being incidental to the argument. Producing that record as a by-product of the close, rather than as a documentation project after it, depends on a model that can hold a long and messy document set in view at once and reason across it rather than merely retrieve from it. Fund agreements are long, side letters contradict each other, and the relevant sentence is frequently three clauses away from the defined term that gives it meaning. We build on Anthropic's Claude for this work because performance on long, internally inconsistent document sets is the thing that determines whether the citation is genuinely right, and because reasoning you can inspect is the difference between a citation an auditor accepts and one they have to verify themselves. It also matters that the model can be connected to the systems of record directly, through the MCP pattern, so that the ledger entry it cites is the live entry rather than a copy someone exported in March.
Speed is a precondition, and we are not arguing against it
The honest concession is that none of this replaces the value of being quick. A slow close is genuinely damaging. Capital account statements arriving eight weeks after quarter end tell an allocator something unflattering about the operation behind them, and the fact that the numbers are eventually well supported does not repair that impression. Speed also has a compounding internal benefit: teams that close quickly have time to look at what the numbers mean, and teams that close slowly spend the following month preparing to close again.
There is a second fair objection, which is that full traceability has historically been expensive, and expensive things get dropped under deadline pressure. That was a legitimate reason to accept a thinner audit trail when constructing one meant a person writing memoranda nobody read. It is a weaker reason once the record is generated by the same process that resolves the exception.
So the argument is not speed versus rigour. It is that speed is becoming the entry requirement and rigour is becoming the thing that still distinguishes one finance function from another.
What the unexplained close costs, on illustrative numbers
To put rough figures on it, and these are modelled rather than observed, take a mid-market platform running three active funds with around a hundred and forty limited partners between them. Assume a three day close, which is excellent, and assume it generates something like sixty substantive follow-up questions a quarter across auditors, operational due diligence reviews and LP finance teams.
Assume, illustratively, that each question costs an average of two and a half hours of controller or chief financial officer time once you include the retrieval, the reconstruction and the internal check before the answer goes out. That is a hundred and fifty hours a quarter, six hundred a year, sitting with the most expensive people in the finance function and arriving in unpredictable bursts.
Now assume traceability cuts the retrieval element rather than the volume of questions, because the questions are mostly legitimate and will keep coming. On the same illustrative basis, if the evidence is already assembled and the answer becomes a review rather than an investigation, the per-question cost falls towards half an hour. The saving is not the point. The point is what happens to the answer itself: it arrives the same day, it is consistent with the answer given last quarter, and it does not depend on which member of staff is still at the firm.
The close stops being the end of the conversation
The finance functions we expect to be admired in five years will not be the fastest ones, because by then the spread between the fastest and the median will be a couple of days and nobody will care. They will be the ones where any figure in any statement can be interrogated immediately, by an auditor, an allocator or a partner, and the explanation comes back with its sources attached and in language the questioner can actually use.
That changes the character of the work. A close that explains itself is not a defensive exercise, it is a foundation. It makes continuous reporting feasible, because a number that carries its own evidence can be released without a review cycle bolted on top. It makes the annual audit shorter. It makes the treatment of an unusual item consistent between one quarter and the next, because the reasoning was written down rather than remembered.
Most of all it moves the finance team back towards the part of the job they were hired for. Assembling the numbers has been the work for a long time because assembling the numbers was hard. As that eases, what is left is explaining what the numbers mean to the people making decisions with them, and that has never been a task anyone wanted to do faster at the expense of doing it well.
