NetSuite Close Benchmarks: Verify the Source and Clock

Executive Summary

This report answers one question for NetSuite finance sponsors: which published month-end close numbers can legitimately be compared with an internal reporting clock, and which should be excluded from a business case? The most widely repeated source is a column published on March 5, 2018 by the chief financial officer of APQC, a nonprofit benchmarking organization ^[1]. It reports a survey of 2,300 organizations: a 6.4 calendar day median ^[1], a top quartile of 4.8 days or less ^[1], and a bottom quartile of 10 or more calendar days ^[1]. The measure runs from the trial balance to the consolidated financial statements ^[1]. These are historical figures, not 2026 results and not NetSuite-specific benchmarks.

A NetSuite article dated December 19, 2018 repeats the bottom-quartile threshold ^[2].

Current numbers are no easier to compare. A vendor-commissioned survey published in October 2026 reports an overall average of 6.2 business days ^[3]; a separate vendor report gives 7 working days ^[4]; and a 2023 community survey reports 8 days without a stated unit ^[5]. APQC itself publishes at least three differently bounded monthly-close measures, one beginning with flash reports ^[6] and one ending at the period-end management report ^[7]. On the NetSuite side, closing a period is a status change ^[8], and the Intelligent Close Manager task-completion figure is a ratio of completed to total tasks, not elapsed days ^[9].

The recommended rule is simple. Accept a comparison only when owner, year, population, statistic, unit, start event, end event, entity scope and counting convention all match; otherwise label it non-comparable or unknown. A fictional example (Hypothetical Example) shows the same sign-off producing 6, 5, 4 or 3 "days" depending on the convention, and a claimed 50% improvement that falls to 33.3% under a like-for-like clock. No verified current APQC median is claimed here, because the current public measure page was not available for review at the time of writing.

Introduction and Background

A close-speed claim is attractive in a business case because it is a single number. A vendor says the average close takes ten days; a project promises three; the sponsor computes a reduction. The difficulty is that every term is a decision: which days, counted how, from which event, to which event, for which entities, according to which survey.

This report is a measurement-certification and citation reference, not a guide to accelerating the close, deploying Intelligent Close Manager, the first close after a go-live, or assembling a general audit-evidence pack. It addresses two separate failure modes. The first is that a published survey statistic can be mischaracterized as it is repeated: a threshold becomes an average, a calendar-day figure becomes a business-day figure, a 2018 sample becomes "current". The second is that the reader's own close timestamp may measure a different event from the one the benchmark measured.

The article is published by Houseblend, a Montréal NetSuite consultancy that implements, integrates and administers NetSuite for venture-backed scale-ups and mid-market companies. That is disclosed commercial context: the publisher has an interest in reporting and process design work. The publisher's own articles are not used as evidence here; claims about third-party figures and NetSuite documentation are cited to the original sources, and items labelled as computed or hypothetical are the publisher's own. No provider is ranked, and the publisher does not appear in any comparison table.

The scope is also bounded by what could be verified. APQC's current public measure page for monthly consolidated statements was not available for review at the time of writing, so no current APQC median, sample, percentile or publication date for that measure is asserted. Where a behavior of a NetSuite account depends on configuration, version or permissions, the report names it as something to test rather than asserting it.

Terms used. Calendar days count every date on the calendar, including weekends. Business days (also called working days or weekdays) count only days in a defined working week after removing a stated holiday list. Start-exclusive counting does not count the start date; end-inclusive counting does count the end date. A quartile threshold is a cut point in the distribution of responses, not a typical value. The clock is the full definition: unit, start event, end event, scope and counting convention.

Key Changes in How a Close Claim Should Be Accepted

Change 1: Treat a benchmark as a ledger record, not a number

A quotable number should carry its provenance with it. The APQC-authored column is a good example of a source that is easy to describe completely. The author's bio identifies the author as APQC's CFO, secretary and treasurer, and describes APQC as a nonprofit benchmarking and best practices research organization. The column attributes the data to APQC's General Accounting Open Standards Benchmarking survey ^[1]. It defines the measure as the cycle time in calendar days between running the trial balance and completing the consolidated financial statements, and it defines cycle time as total elapsed time from the beginning of the process to the end ^[1]. The page does not state a survey year; the only date it supports is the March 5, 2018 publication date.

Because the author is an APQC officer writing about APQC data, the column is close to a primary source for the figures. A vendor article may still be useful as corroboration, but it does not become the originator. NetSuite's continuous-accounting article, bylined to a NetSuite solutions center and dated December 19, 2018, restates the same sample size ^[2], the same definition ^[2] and the top-quartile figure. It does not carry a link to APQC, so a reader must trace the originator independently. For the ledger below, the originator is APQC, the earliest dated statement is the 2018 column, and the NetSuite page is a dated secondary restatement.

Statistic type matters as much as ownership. The median is the value with half the observations on each side ^[10], and the mean is the sum divided by the count, the quantity most commonly called the average ^[10]. Percentiles split ordered data into hundredths ^[11]. APQC's own FAQ describes its 25th percentile as the performance level at which 25% of all responses fall below ^[12]. For a duration metric, where lower is better, that cut point is the "top quartile" boundary; the bottom quartile is the other tail. A bottom-quartile threshold of "10 or more" therefore says that a quarter of respondents were that slow or slower. It does not say that the typical organization takes ten days.

Each ledger entry therefore records the following fields:

• Originator: the organization that owns the research, not the page that repeats it.
• Publication date and survey period: the dated statement on the page, and "not stated" when the survey year is absent.
• Sample and population: count, respondent type, size band and industry cut.
• Statistic type: median, average, quartile threshold or share over a threshold.
• Unit: calendar days, business days or working days.
• Start and end events: as worded by the originator.
• URLs and verification date: original, secondary, and the date each was opened.
• Status: approved, directional, non-comparable or unknown, with access-restricted pages recorded as unknown.

Two further ledger fields are easy to forget. Quartile cut points depend on the interpolation method, and the NIST handbook notes that the R6, R7 and R8 methods give fairly similar but not exactly identical results ^[11] and that all the methods it describes are used in practice ^[11]. And a peer cut is part of the definition: APQC's materials describe peer groups by industry, region and revenue ^[13], with industry peers segmented into clusters ^[12]. A figure without its peer cut is a different statistic from the same figure with one.

Change 2: Replace "close" with named events

The word "close" covers at least seven separable events in a NetSuite finance operation: the period end, the first agreed trial balance, entity-level reporting sign-off, completion of group consolidated statements, completion of the NetSuite checklist, the period lock or close, and board-pack release. Accounting standards treat several of these as distinct dates. IFRS defines events after the reporting period as those occurring between the end of the reporting period and the date the financial statements are authorised for issue ^[14], and requires disclosure of that authorisation date ^[14]. US GAAP has its own issuance dates; FASB's Accounting Standards Update 2010-09 distinguishes statements that are "issued" from those "available to be issued" ^[15]. The auditing literature likewise separates the as-of date from the period-end reporting process, which includes preparing the financial statements and related disclosures ^[16].

NetSuite's documentation draws a similar line between system state and financial reporting. Oracle defines closing an accounting period as changing its status from Open to Closed ^[8], and says a Closed status indicates that posting transactions for the period have been completed ^[8]. It recommends closing as a final step after accounts review ^[8]. Locking is a distinct, earlier control: it prevents users without override permission from posting to the period ^[17]. Nothing in these pages equates a Closed or Locked status with the agreed consolidated financial statements.

The same underlying work can therefore be described by several valid clocks; the invalid practice is comparing a number from one clock with a benchmark from another. APQC's own catalog illustrates the point: one public measure counts calendar days "including weekends" from the initial monthly flash reports ^[6], another runs from the initial business-entity trial balance to the period-end management report for senior management ^[7], and another measures the gap between completed consolidated statements and the release of earnings ^[18]. All three are "close" metrics. They have different start and end events, so their medians should not be placed on one chart.

Change 3: Replace conversion ratios with counted dates

Reports that mix units are common. Rand Group's guide, dated June 1, 2026, restates the APQC definition in calendar days ^[19], and in the same article gives its own guidance that a healthy close should take between three and six business days ^[19], with larger or more complex organizations needing six to ten business days ^[19]. The guide labels its own guidance as business days and the APQC measure as calendar days. The risk arises when a reader or a downstream summary places "10 calendar days" and "10 business days" in one sentence. For example, ten calendar days after Friday, July 31, 2026 is Monday, August 10, which is six weekdays under a start-exclusive Monday to Friday count (our computation), and the relationship shifts with the weekday on which the period ends. No fixed percentage converts one to the other, so none is used in this report.

The right approach is to count actual dates under a written convention. Python's date arithmetic returns an exact elapsed difference ^[20], which is a start-exclusive calendar count. NumPy's busday_count includes the begin date and excludes the end date ^[21], accepts a weekmask defining which weekdays are valid ^[21], and takes a holiday list of dates to treat as invalid ^[21]. Excel's NETWORKDAYS excludes weekends and any listed holidays ^[22], and Microsoft's worked example counts the workdays between a start and an end date ^[22]. Our own computation of that example, 110 for October 1, 2012 to March 1, 2013, matches counting both endpoints; this is an inference from the example, not a statement on the page. The practical point stands: two common tools can return different counts for the same two dates, so the worksheet must record which was used.

Holiday calendars must be specified by jurisdiction. The US Office of Personnel Management list is established for federal employees only ^[23]; in 2026, because July 4 falls on a Saturday, the preceding Friday is treated as the holiday ^[23], and Labor Day falls on Monday, September 7 ^[23]. In Canada, the Canada Revenue Agency lists the Civic Holiday as Monday, August 3, 2026, excluding Quebec ^[24], and notes that provinces and territories have their own holidays ^[24]. British Columbia lists B.C. Day on that date ^[25], Saskatchewan lists Saskatchewan Day ^[26], Ontario does not require employers to give the first Monday in August ^[27], and Alberta lets employers choose to recognize additional days ^[28]. In the United Kingdom, GOV.UK lists a Summer bank holiday on 3 August for Scotland ^[29] and on 31 August for England and Wales ^[29]. A single "business day" definition for a multinational group therefore does not exist; each entity's calendar must be listed.

Change 4: Evidence each timestamp, and say when no system report exists

Each event in the clock needs an owner of the evidence. Some events are evidenced by NetSuite state changes. Others are evidenced only by a controlled register kept outside NetSuite, such as a controller's sign-off log. The distinction is the heart of the acceptance test.

On the system side, Oracle documents that the Period Close Checklist gives access to system notes and user notes related to period close ^[30]. System notes capture the date and time of a change and who made it ^[31], and they cannot be edited by any user, script or app ^[31]. These are strong evidence of when a status changed. But the details limit what they prove. Viewing a checklist task for the first time logs a Create entry in System Notes ^[30], so a note's existence does not show completion. The checklist's Modified By field for a subsidiary lock changes only when all subsidiaries have been locked ^[30]. And the Quick Close function only marks tasks as completed ^[8], while the audit trail still records every changed task.

Scope adds more variants. In NetSuite OneWorld, periods can be locked at the subsidiary level ^[32], and a lock task shows as partially done when at least one subsidiary is locked ^[32]. With Multi-Book Accounting and the extended period close process, periods for each accounting book can be closed and reopened individually ^[33]; each book has its own close status flag while the status accessible through SuiteScript remains shared ^[34]. The Intelligent Close Manager portlet displays data only for the primary accounting book ^[35], and its system-generated tasks and KPIs refresh every hour ^[35]. Portlet data and period-lock status are useful operating signals. None is a days-to-close metric, and none records controller sign-off of the consolidated statements.

Documented saved-search fields include a Date Closed On field described as the date when the period was closed and an All Locked field. Whether a date-only field is adequate, which time zone it uses, and whether an equivalent column is exposed through SuiteQL or SuiteAnalytics in a given account and release are not documented in the pages reviewed for this article and must be tested in the actual account. The same caution applies to any claim of a universal native elapsed-close metric: none is documented in the pages reviewed.

Change 5: Treat improvement as a like-for-like calculation

A before-and-after percentage is valid only when the clock and scope are identical and the baseline is nonzero. The formula is (baseline duration minus new duration) divided by baseline duration. Without the identity condition, the result measures a change in definition. Three common sources of silent definition change are:

• Reopened periods: the end event moves when a period is reopened and reclosed.
• A changed start event: period end versus first agreed trial balance.
• A changed entity scope: subsidiaries, books or currencies added or removed.

Reopening is a documented NetSuite action. When a period is reopened, any later closed periods are automatically reopened ^[36], and the justification entered is saved in a user note ^[36]. The clock must therefore state whether the end event is the first sign-off or the last sign-off after any reopening, and the change log must record which one each period used. Mixing the two across months will move the series without any change in operations.

External clocks add a third dimension. SEC Form 10-Q sets filing deadlines of 40 days after quarter end for large accelerated and accelerated filers ^[37] and 45 days for all other registrants ^[37], and a report otherwise due on a weekend or federal holiday is due the next business day ^[38]. These are calendar-day regulatory deadlines for quarterly filings, not internal close times, so a filing-deadline clock and an internal reporting clock must not share a label.

Implementation Considerations and Process Changes

The following is a clock-definition acceptance worksheet. It is designed to be completed once per metric, approved by the controller, and versioned. It is a measurement control, not an improvement plan.

Worksheet A: define the clock

• Metric name and owner: one named finance owner accountable for the definition.
• Start event: period end, first agreed trial balance, or another named event, with the evidence source.
• End event: entity sign-off, group consolidated statements approved, NetSuite lock or close, or board-pack release, with the evidence source.
• Unit: calendar days or business days, never "days".
• Counting convention: start-exclusive or start-inclusive; end-inclusive or end-exclusive; tool used (for example date subtraction, busday_count or NETWORKDAYS).
• Holiday calendar by entity: jurisdiction, year and published source for each entity, with the 2026 dates actually applied.
• Time zone: the zone in which the timestamp is recorded and the zone in which the date is derived.
• Scope: subsidiaries, accounting books, currencies, and whether eliminations and consolidation adjustments are included.
• Cutoff and version: the report version that counts as final, and whether later revisions reset the clock.
• Reopen handling: first sign-off, last sign-off, or both reported.

Time zone is not a technicality. NetSuite documents that the chosen time zone affects events and the data returned by reports and searches for date fields with timestamps ^[39], and that report results for those fields are based on company or user time zone rather than system time zone ^[40]. Classic system notes display date and time in the company time zone. Time zones are identified by IANA identifiers in the standards ecosystem ^[41]. A sign-off at 11:30 p.m. Eastern on August 6 is already August 7 in UTC; if one report derives dates in UTC and another in the company zone, the same event lands on different days.

Worksheet B: acceptance tests

Run these before a number enters a target, a service-level agreement or an automation business case.

• Same clock across periods: start, end, unit and convention are identical in every period of the series.
• Entity and book agreement: subsidiaries and accounting books included in the baseline equal those in the new measure.
• Reopen consistency: reopened periods and revised financials are treated by the same rule in every period, and the change log lists each instance.
• Dated holiday test: a recorded example, such as the one below, passes when recomputed by a second person under the written convention.
• Time zone test: the same sign-off timestamp yields the same date in every report and extract used.
• Waiting is not labor: elapsed time includes queues and waiting. The 2018 APQC column defines cycle time as total elapsed time ^[1], so a reduction in active labor hours is not evidence of a shorter elapsed close.
• Draft is not final: a fast first draft is never counted as final approved reporting.
• Survey match: the public survey's year, population, statistic type and percentile are identical to those quoted.

Worksheet C: approve, qualify or exclude

• Approved comparison: every field matches; the comparison can enter the business case with its ledger entry attached.
• Directional only: one field differs in a known, documented direction (for example, business days versus calendar days with a known holiday list); use the figure as context and state the difference.
• Non-comparable: start or end events differ, or the statistic type differs; exclude from the target and from any improvement percentage.
• Unknown: the field cannot be verified (for example, a paywalled report or a page that cannot be opened); record the status and do not fill the gap with a current-sounding date.

Benchmarking, as ASQ describes it, measures products, services and processes against those of leaders ^[42], and requires that the organization's own processes be clearly understood and under control ^[42]. For scale-ups, the practical implication is to settle the definition while the entity count is small; a company that adds subsidiaries, books or currencies later can otherwise find that its series quietly changed scope.

Data Analysis and Evidence

The provenance ledger

Table 1 below records the benchmark claims reviewed, with the fields the worksheet requires. Statuses reflect the sources reviewed for this article. "Not stated" means the opened page did not state the field. "Unknown" means the page was not reviewed for this article, and no value is supplied.

Table 1 shows that no row is directly comparable with another. Even the three APQC-derived rows differ in vintage, sample and, for the other APQC measures, in endpoints. The recent survey rows use business or working days and report averages or bands, while the 2018 column uses calendar days and quartiles. The dates and unit fields are the first things to check.

Reading the other surveys

Several additional published figures provide context, with limits. A 2022 analyst perspective from ISG's research group (formerly Ventana Research) reports that the percentage completing the monthly close in seven or more days was largely unchanged at 41% versus 40% ^[44]. A 2025 SAPinsider report describes a narrower phase, noting that the average company takes more than nine days for statement review and finalization ^[45]. Gartner lists a "Days-to-Close Benchmarks" document that varies by company size, but only the abstract is public ^[46], so its figures are Unknown for this report.

For "days to close by company size", the reliable public evidence is thin. Vena's article reports subgroup averages by number of entities ^[3], and the survey is unweighted and vendor-commissioned ^[3]. APQC's annual-close blog reports that top performers complete the annual close in 10 days or less against a median of 18 days ^[47], but that is an annual, not a monthly, measure. No independent monthly benchmark by company size was verified for this article, so a size-adjusted target should be built from the organization's own history until a definition-matched peer cut is obtained.

The evidence-to-clock matrix

Table 2 maps each event to the evidence that can establish its timestamp. "Supported report" means a documented NetSuite capability that should be tested in the target account; "manual register" means a controlled log maintained by finance.

Table 2 is deliberately conservative. Four of the eight table rows (first agreed trial balance, entity sign-off, group consolidated completion and board-pack release) have no documented NetSuite state and need a manual register. Closed status shows that posting is finished, not that the numbers were approved.

Case Studies and Real-World Examples

A Dated Calendar Example (Hypothetical Example)

This example is entirely fictional. It does not represent a real client, a published benchmark or any observed company. Its purpose is to show how counting conventions change a result. Assume a July period ending Friday, July 31, 2026, and a controller-approved consolidated reporting package on Thursday, August 6, 2026. The first agreed trial balance is produced on Monday, August 3. Only Variant D assumes that August 3 is a holiday in the entity's jurisdiction under a stated calendar, for instance British Columbia, where B.C. Day falls on that date ^[25]; the other variants assume no holidays.

Table 3 below lists the same events under different conventions. The calendar rows are straightforward date subtraction ^[20].

The six results come from one fixed sign-off date. Variant B matches NumPy's busday_count call from July 31 to August 6 only because both endpoints are weekdays; the library includes the begin date and excludes the end date ^[21], and the two methods diverge when an endpoint is a weekend. Variant C follows the inclusive-endpoints reading of Excel's NETWORKDAYS example described above ^[22]. Variant D uses a holiday list as the library permits ^[21].

Now suppose a fictional baseline month was measured as Variant A, 6 calendar days. A fictional later month, after a hypothetical automation project, reaches sign-off 4 calendar days after period end under the identical clock. The like-for-like improvement is (6 minus 4) divided by 6, or 33.3%. Suppose that in the later month the first agreed trial balance is produced one calendar day after period end. If its result were instead quoted from that event as 3 days, the claimed improvement becomes (6 minus 3) divided by 6, or 50.0%. The additional 16.7 percentage points come entirely from the changed start event, not from any faster work. Under the acceptance worksheet, the second comparison is Non-comparable and excluded. These figures are arithmetic on fictional dates and must not be read as an empirical improvement rate.

Time zone adds a last wrinkle. If the sign-off timestamp were August 6 at 11:30 p.m. in Toronto (UTC minus 4 hours in summer), a UTC-based extract would date it August 7 at 03:30. A one-day shift appears with no change in operations, which is why Worksheet A requires the derivation zone.

Implications and Future Directions

For finance sponsors, the main implication is that a measurement contract is cheaper than a disputed benchmark. A one-page ledger entry per cited statistic, plus a signed clock definition, lets the finance team defend a target to an audit committee, a board or a vendor. It also protects against a more mundane risk: a stakeholder reading a headline number out of context. The 2018 column itself reminds readers that cycle time is the total elapsed time, not the active effort.

For vendors and implementers, the implication is that product completion scores and reported close times answer different questions. Intelligent Close Manager reports a ratio of completed tasks for the primary book on an hourly refresh. A project sponsor who approves automation on the strength of a task-completion rate has approved a different outcome from one who approves a reduction in the elapsed time to approved consolidated statements. Both are legitimate goals; they should be written down as separate goals.

APQC states that submissions pass a logical and statistical validation process ^[12], but also that participants must complete at least 80 percent of the metric questions to receive a report ^[12], so the repository composition is self-selected and may change by vintage. This report does not evaluate that composition; it records that the question exists.

Two items remain open. First, a current APQC cut for the monthly consolidated statements measure, by industry and revenue, would replace the historical anchors; it is not verified for this article. Second, a tested NetSuite extract of period close timestamps, including the time zone of any date field and the behavior with Multi-Book Accounting, should be documented per account and release. Existing improvement and dashboard guidance becomes useful only after the measurement has been approved.

Frequently Asked Questions (FAQs)

What is a financial close benchmark definition?

A usable definition names the owner, year, population, statistic, unit, start event, end event, scope and counting convention. The 2018 APQC column gives a compact example.

What is the average month-end close time?

No single answer exists. Recent vendor surveys, some with unclear publication dates, report 6.2 business days, 7 working days ^[4] and 8 days with no stated unit ^[5], using different samples and definitions. APQC's historical median was 6.4 calendar days from trial balance to consolidated statements ^[1]. These are not interchangeable.

What is the NetSuite financial close time?

NetSuite documents period status and task state, not a universal elapsed-close metric. Closing is a status change, and the Intelligent Close Manager task-completion percentage divides completed task items by total tasks. A NetSuite close time is whatever the clock in Worksheet A defines, evidenced by the sources in Table 2.

How should a team benchmark its financial close process?

Document its own process first, as ASQ's benchmarking guidance recommends ^[42]. Then complete Worksheets A and B, build a ledger entry for each external figure, and approve, qualify or exclude each comparison under Worksheet C.

How do close days compare by company size?

Public monthly evidence by size is limited. A vendor-commissioned survey reports subgroup averages by entity count, and APQC's annual-close blog reports annual (not monthly) medians of 10 days for organizations under $100 million in revenue and 23 days for those with $1 billion to $5 billion ^[47]. Neither is a verified monthly, definition-matched benchmark for a specific NetSuite company size.

What does the NetSuite period close checklist show?

It gives access to system notes and user notes ^[30], and status icons show whether each task is completed, partially done or not started. Checklist tasks are listed in the order in which they should be completed ^[30], and closing waits until all previous periods are closed. The checklist shows state, not approved reporting.

Are business days or calendar days correct?

Neither is universally correct; the clock must state which. APQC's measure is in calendar days, while Rand Group's own healthy-close guidance is in business days. Convert only by counting actual dates with a stated holiday list, because holidays differ by jurisdiction, for example August 3, 2026 ^{[24] [29]}.

Does a quarterly filing deadline define the close?

No. Form 10-Q deadlines are calendar-day filing limits measured from quarter end ^[37], a different clock from the internal reporting date.

What are practical month-end close practices for scale-ups?

From a measurement viewpoint: fix one written definition early, record entity and book scope as it grows, keep a manual register for sign-off, and log each reopening with its justification. Speed improvement techniques are outside this report's scope.

Conclusion

The headline close numbers in circulation come from different owners, years, populations, statistics and units. The best-documented historical anchor, the March 2018 column by APQC's CFO, reports a 6.4 calendar day median, a 4.8 day top quartile and a 10 or more day bottom quartile for 2,300 organizations on a trial balance to consolidated statements clock. It is a reliable record of what it says, and it is historical. Any later restatement should be checked for statistic type, unit and population before it is used. The sensible response is a ledger, not a debate.

On the reader's side, NetSuite supplies strong evidence for system state changes and weaker, or no, evidence for approved reporting. Status changes are logged and immutable in system notes, but a Closed status is a posting status and the task-completion KPI is a ratio. Agreed financial statements require a manual register of sign-offs unless a tested, supported report is shown to capture them.

The decision rule is therefore concrete. Approve a comparison only if every ledger field and every clock field matches; qualify it if the difference is known and documented; exclude it from the business case if events or statistics differ; and record Unknown when a source cannot be opened. By that rule, this report approves no current APQC figure, because none was verified, and treats every vendor-survey average as context rather than a target. Reviewers should consult a qualified accounting professional on how these definitions interact with their own reporting obligations.