Ask ten finance chiefs how long their close takes and you will get ten numbers. Ask what those numbers measure and the conversation stops.

The most cited benchmark in the field comes from APQC, which puts the median monthly close at 6.4 days across roughly 2,300 organizations. Top quartile is 4.8 days or less. Bottom quartile is ten or more.

Useful numbers. They are also calendar days rather than business days, and they run from the moment you execute the trial balance to the moment consolidated statements are complete. Not from period end. Not to anything a board, a lender or a regulator ever sees.

Most finance functions measure something quite different, and then compare it to this.

With year-end thirteen weeks out, it is worth being precise about what you are actually trying to shorten — because the answer determines whether the fix costs money or costs nothing.

Calendar days or business days. APQC counts calendar days, weekends included. Most finance teams count business days. Over a typical close that difference is worth two or three days on its own — the same process, described two ways, with a gap that exists entirely in the definition and none of it in the work.

Where the clock starts. APQC starts at the trial balance. Most teams start at period end. In between sits everything that has to arrive before the trial balance is meaningful: AP cutoff, payroll, bank feeds, inventory counts, subsidiary submissions. For a single-entity business that might be a day. For a multi-entity private company waiting on managers who have other jobs, it is routinely four or five.

Where it stops. Trial balance closed. Consolidated statements complete. Reviewed. Approved by the audit committee. Released. Filed. Each of those is a real endpoint, each is days apart from the last, and people quote whichever one flatters.

And which close. APQC’s annual median is 18 days against 6.4 for a month, with slower performers at 35. Organizations under $100 million in revenue post a median annual close of 10 days; those between $1 billion and $5 billion take 23. Benchmarking a year-end against a monthly figure is the most common error in the category, and it is always flattering in the wrong direction.

Here is where the confusion has consequences.

A large accelerated filer has 40 days from quarter end to file a 10-Q. APQC separately measures the gap between completing quarterly statements and releasing earnings, and puts the median at 15 days.

So a public company can accurately say its close takes six days and just as accurately file on day 38. Both numbers describe the same quarter. The first is an accounting process. The second is a reporting process, and most of the distance between them is review, drafting, legal, audit committee scheduling and the mechanics of the release.

Compressing the close by two days does nothing for the filing date if the bottleneck sits in the second half of that chain. A good many close acceleration projects have delivered exactly that: a faster close and an identical filing calendar.

The private company version is the same problem with different dates. Trial balance, then statements, then the board pack, then the covenant compliance certificate — which is frequently not due for 30, 45 or 60 days after quarter end.

A private company closing in twelve days and reporting to its lender on day 45 has more than a month of slack that nobody is measuring. Which is worth saying plainly: the reason many private closes take twelve days is that nothing bad happens on day twelve. The absence of a deadline is doing exactly what the absence of a deadline does.

That is not a criticism of private finance teams. It is an observation about incentives. Public companies did not get faster because they are better. They got faster because a statute made the date real.

Once you know which link you are trying to compress, the interventions divide cleanly, and they are not the ones vendors lead with.

If the delay is before the trial balance, the fix is sequencing and cutoff discipline. Hard AP cutoffs. Accrual estimates for anything below a materiality threshold rather than waiting for the invoice. Subsidiary submission deadlines that are enforced rather than requested. None of this requires software, and all of it requires someone senior to say no to a late invoice.

If the delay is in the trial balance to statements window, the fix is usually reconciliations and the chart of accounts. Accounts that are reconciled monthly do not produce surprises at year-end. Accounts that have never been reconciled produce all of them at once. And a chart of accounts that requires manual mapping to produce a consolidated view will require it every single month, forever, until somebody restructures it.

If the delay is after the statements are done, no accounting improvement will help. The constraint is review capacity, audit committee calendars, legal turnaround, or the availability of one person who reads everything. That is an organizational problem and it is solved by scheduling, not by automation.

The benchmark studies claim automation delivers a 30% to 50% reduction in close cycle time. Those figures come largely from vendors and their research partners, and they are probably true for the subset of companies whose bottleneck is manual reconciliation. For a company whose real delay is a controller waiting on four subsidiary submissions, the software will reconcile faster and the close will finish on the same day it always did.

And a word on AI specifically, because it is being sold hard into this problem. The tasks AI handles best right now are explanatory: drafting variance commentary, producing a first pass at the flux analysis, assembling board materials from finalized numbers. Those sit after the close, not inside it. The work that actually delays a close is waiting for inputs that have not arrived and resolving items that require judgment, and no model makes a subsidiary controller submit on time.

There is a second-order point worth holding. We reported last week that only 7% of finance leaders describe themselves as very confident interpreting AI outputs, and 46% have no formal AI governance at all. A close that finishes two days earlier on numbers nobody can verify is not two days faster. It is two days earlier and less certain, which in this particular process is not a trade most CFOs would make deliberately.

Our take: before buying anything, spend one cycle timestamping the close. Note when the period ended, when the trial balance ran, when statements were complete, when review finished, and when the last person who was waiting stopped waiting. Five timestamps. Then look at where the days actually sit. Most finance functions discover the answer is not where they assumed, and a meaningful number discover the largest single block is a queue rather than a task.

Every customer signal is captured in one shared context layer, always current and compounding. Agents and workflows build pipeline, chase every buying signal, and move deals forward, an always-on revenue engine running alongside your team.

With Attio, you’ll get:

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?

Get started now

Last week we asked what rate your 2027 plan assumes, after the Fed’s first increase since 2023 and projections showing no cuts next year.

Forty of you answered. Nineteen — just under half — are carrying a rate below today’s, meaning a cut or two next year. Ten have updated to at or above today’s level. Nine would have to check. Two have not built 2027 yet.

Set that against what the Fed published four days earlier. Its own median projection for the end of 2027 is 4.1%, identical to its 2026 median and implying no cuts at all. The futures market is pricing higher still — and the odds of another increase at the October meeting, five weeks out, have now firmed to around 70%, from just over half a week ago.

So roughly half of the finance functions reading this are planning against a rate path that neither the central bank nor the market currently forecasts. That is not necessarily wrong — plans get built on judgment, and the Fed has been wrong before. But it is a position rather than a default, and the nine who would have to check do not yet know which one they are holding.

Four things every CFO needs to know this week.

· CFO confidence rebounded, and the detail is where it gets interesting. Deloitte’s Q3 CFO Signals, published this week, polled 200 North American finance chiefs at companies above $1 billion in revenue. Confidence recovered. Revenue growth expectations rose to 4.6% and capital expenditure to 4.3%, both modestly ahead of last quarter. But expectations for earnings, dividends, domestic wages and domestic hiring all declined. Read those together and the picture is a cohort planning to spend more, sell more, and employ fewer — which is a coherent position, and not an optimistic one.

· Rivian’s controller followed its CFO out the door. Corporate controller Jon Franklin announced his departure late last week, about a month after long-time CFO Claire McDonough left for GE Vernova. A VP of finance is serving as interim while a search runs. The most useful observation came from an anonymous commenter rather than an analyst: the finance organization that carries a company through startup, IPO and capital raising is not necessarily the one it wants for scale and profitability. That is a real argument, and it applies well beyond one automaker — the skills that get a function through a transaction are not the skills that run it afterward.

· Oil and yields both moved on the strait. The US rejected Iran’s proposal to reopen the Strait of Hormuz, and crude and Treasury yields rose together. We flagged the freight consequence last month: diesel and fuel oil were both running more than 50% above year-ago levels. If your 2027 plan assumes energy normalizes, this is the second month in a row the assumption has been tested, and the Fed has now told you it is not planning to cut into it.

· The consumer test we flagged in August prices this week. Oura, the Finnish smart-ring maker, is marketing 50 million shares at $40 to $44 on Nasdaq, a fully diluted valuation around $15.6 billion against $11 billion in a private round last October. Revenue reached $1.21 billion for the nine months to June 30, up roughly 74%, on a $5.99 monthly membership attached to the hardware. Reuters called it a test of appetite for consumer technology after a tepid start to the fall season, which is the same test we described two weeks ago. Note the structure: of the 50 million shares, the company is selling 13.5 million and existing stockholders 36.5 million. Nearly three-quarters of the deal is insiders taking money off the table, and Oura sees none of those proceeds. That is a different signal from a company raising capital to build something.

1. What are your five timestamps? Period end, trial balance, statements complete, review complete, and delivery to whoever was waiting. One cycle of data. Until you have it, every conversation about close speed is about a number nobody has defined.

2. Which endpoint are you actually judged on? For a public filer it is the filing date, and the close is one input. For a private company it is usually the board meeting or the covenant certificate. Shortening a link nobody is waiting on produces no benefit, and it is the most common form of wasted effort in this area.

3. How many accounts are reconciled monthly? Not how many should be. How many are. The accounts that skip months are the ones that generate the year-end surprises, and year-end is thirteen weeks away.

4. What is the longest single queue? Not the longest task. The longest wait — the period where the file sits with someone who has not looked at it yet. In most closes the largest recoverable block of time is a queue, and queues are solved by scheduling rather than by effort or software.

5. If you had three fewer days, what would you do with them? A serious question. If the answer is that the numbers would reach the board at the same meeting anyway, the close is not your constraint and the project should be something else.

SPONSOR THIS NEWSLETTER

Your name, in front of the CFOs and senior finance leaders who read The CFO Desk every Tuesday. The primary slot is available for Q4. Write to [email protected].

The close is not one number. It is a chain of handoffs, and the people at each end of it are measuring different things and comparing notes as though they are not.

Six days and thirty-eight days can describe the same quarter. So can twelve days and forty-five. The useful question is not which number is right. It is which link has someone waiting at the end of it.

One question before you go: where is AI actually earning its place in your close?

One click. And if you have a view on why, reply — we read everything.

The Editors hold no position in any company mentioned.

— The CFO Desk
[email protected] · thecfodesk.com
CAPITAL · FINANCE · LEADERSHIP