A drug company pays hospitals, labs and research firms to run its trials, and most of them send their bills weeks or months after the work. The company still has to show that cost in the quarter the work happened. So at each quarter end its accountants work out, study by study, what has been done and not yet billed, and book that estimate. The estimate is the clinical trial accrual.
This guide is for anyone who builds, reviews or audits that number at a drug, biotech or device company: clinical accountants, R&D controllers, study managers who answer accrual questions, audit liaisons, and leaders deciding whether to automate the work. No accounting background needed.
Imagine you are the clinical accounting lead at a mid-size biotech with eight studies running. It is Wednesday, October 7, 2026, at 7:55 am, and your journal entries go into the books tomorrow at noon. A research firm, Tessaly Research, sends its estimate for a late-stage (Phase 3) lung cancer study: $81,300 of site costs done and not yet billed at September 30.
You work it out yourself. The visit records show 4,410 patient visits, each priced at its hospital's budget: $5,402,300 earned. Invoices so far: $5,290,100. The difference is $112,200, which is $30,900 more than Tessaly's figure. Tessaly dated its estimate September 26 and assumed 31 screening visits for the month. The records at September 30 show 43.
Tessaly's estimate · dated Sep 26
Screening visits
31
After that date
Not counted
Site costs unbilled
$81,300
Visit records · at Sep 30
Screening visits
43
Extra, at $2,575
12 = $30,900
Site costs unbilled
$112,200
$81,300 estimated$112,200 bookedwith the reason written down, and the vendor asked to confirm.
Careful accrual work catches twelve patients screened in the quarter's last four days, before the auditors or next quarter's invoices do.
How clinical trial accruals work: earned minus billed, the vendor's estimate, the auditor and the rules
What the number is, what it is made of, who tests it and how it goes wrong, explained without selling anything. If accruals are new to you, start here.
01 · In plain words
A clinical trial accrual in plain words: the work happened in September, the bill comes in November
Picture a trial testing a new medicine at 46 hospitals. In September, a patient comes in for a screening visit. A nurse draws blood, a doctor examines the patient, and the hospital records it. The drug company running the trial, called the sponsor, now owes that hospital a fee. Nobody has sent a bill yet, and nobody will for a while. The hospital usually bills the research firm managing the trial, the firm collects a month of such costs, and its invoice reaches the sponsor weeks later.
Accounting rules ask the sponsor to show the cost in the period the work was done, whenever the bill arrives.1 So when September 30 comes and the company closes its quarter, its accountants have a problem. They know what has been billed. They have to work out what has been done. The difference between the two, work done and not yet billed, goes on the books as an expense and as money owed. That entry is the clinical trial accrual.
One study can have thousands of visits, each priced by that hospital's own budget, plus a fee each time a monitor visits to check the site's records, a management fee earned month by month, lab kits, imaging reads, patient travel and fee changes signed halfway through. A company with eight studies does this eight times, every quarter, against a deadline. The research firm also sends its own estimate of the same number, and the two rarely match.
Then the invoices arrive and show how good the estimate was. If the company guessed low, next quarter carries the extra expense. If it missed something large, as one public company reported in 2024, past financial statements may have to be corrected, which is called a restatement.2 That is why auditors spend time on a number most people outside finance have never heard of.
One screening visit, from the hospital to the audit file
The contract sets the priceThe sponsor signs a work order with its research firm, and each hospital signs a budget: $2,575 for a screening visit, in this guide's example.
The patient comes inOn September 28 the visit happens, and the hospital, called the site in trial work, enters it in the electronic trial record. Money is now owed, and no bill exists.
The research firm estimatesTessaly Research dated its estimate September 26, so this visit is missing from it.
Finance builds the accrualAt the cut-off, the date the count stops, finance counts the visits, prices them, subtracts what was billed and books the gap.
The invoice proves it, laterThe entry is undone on October 1, the real invoice lands, and the team can compare what it estimated with what came in.
Step 4 is the subject of this guide. The people, companies and figures in this example are fictional.
Four things about trial accruals that surprise newcomers
The research firm's number is only a starting point
One audit and advisory firm warns that biotechs may take a research firm's cost figures "at face value" and still have to form their own view for accounting and audit.3 The firm's estimate is dated before quarter end, so it can miss the last days.
A small balance can still be a critical audit matter
A critical audit matter is one an auditor describes in its report because it took especially challenging, subjective or complex judgment.4 One company's auditor named its clinical trial accruals and related prepaid expenses a critical audit matter for 2024, citing management's judgment in estimating progress and the audit effort it took. The balances were $0.7 million of accruals and $0.5 million of prepaid expenses.5
Money paid early is an asset
A non-refundable fee paid up front for future trial work stays on the balance sheet and turns into expense as the work is done.6 In this guide's fictional example, $3.1 million paid on signing is still in prepaid more than two years later.
The cost of the work is often the smaller argument
The argument is often over timing: which quarter a cost belongs in. Timing is what the public restatement this guide describes turned on: about $3.9 million of clinical expense under-accrued in 2022.2,7
Four terms on every accrual schedule: earned, billed, accrual, prepaid
Earned to date
The value of the work done so far under a contract, whatever has been invoiced.
Billed to date
What the vendor has invoiced so far, as recorded in the company's finance system.
Accrual
Earned above billed: work done, no bill yet. An expense and a liability.
Prepaid
Billed above earned: paid ahead of the work. An asset until the work happens.
Why trial bills lag the work, and what earned minus billed measures
In shortEach line of a trial contract has a value of work done and an amount invoiced. Earned minus billed gives an accrual when positive and a prepaid when negative. The lag comes from how trial money travels: from a visit, through a site and a research firm, to an invoice that covers a month already gone.
Most sponsors hire a contract research organization, or CRO, to run some or all of a trial. US rules let a sponsor transfer its trial duties to a CRO in writing.8 ICH E6(R3), the global good clinical practice guideline, says the sponsor keeps ultimate responsibility for what it transfers and should oversee it (3.6.6, 3.6.9), and that the financial aspects of a trial should be documented in an agreement between the sponsor and the investigator or institution (3.5).9 The money follows those agreements, and each one bills on its own clock.
A visit in late September reaches the sponsor's books by estimate, because the invoice comes after the close
1Visit happens
A patient is screened at a hospital.
SiteSep 28
2Visit recorded
The site enters it in the electronic trial record.
Site staffDays later
3Vendor estimate
The CRO's own estimate, dated before the quarter ended.
CROSep 26
4Quarter closes
Finance counts, prices and books the unbilled work.
Clinical accountantOct 8
5Invoice lands
The CRO bills the month in arrears.
CRO, accounts payableLater in Q4
Fictional example from this guide · Dates from this guide's fictional close. The CRO's estimate is dated before the visit, and that is where the example's gap comes from.
Show the numbers
Step
What happens
Who
Time
Visit happens
A patient is screened at a hospital.
Site
Sep 28
Visit recorded
The site enters it in the electronic trial record.
Site staff
Days later
Vendor estimate
The CRO's own estimate, dated before the quarter ended.
CRO
Sep 26
Quarter closes
Finance counts, prices and books the unbilled work.
Clinical accountant
Oct 8
Invoice lands
The CRO bills the month in arrears.
CRO, accounts payable
Later in Q4
Accrual or prepaid: the same subtraction gives both
Take one line of a contract at the quarter's end. Work out what the vendor has earned. Subtract what it has invoiced. A positive answer is an accrual: the books show more expense and more money owed (debit R&D expense, credit accrued liabilities). A negative answer is a prepaid: the sponsor has been billed for work still to come, and the difference sits as an asset until the work happens.
Do this by line, because netting a study's lines together hides both. In this guide's example one study shows $1,675,048 of accruals and $3,321,570 of prepaid at once, a net of −$1,646,522, and that net figure alone would tell a reviewer almost nothing.
Non-refundable advances have their own rule. CROs commonly ask for money before work starts. ASC 730 is the US accounting standard on research and development. Deloitte's life sciences accounting guide summarizes ASC 730-20-25-13, 730-20-25-14 and 730-20-35-1: non-refundable advance payments for future R&D are deferred and capitalized, meaning held as an asset, then "recognized as an expense as the related goods are delivered or the related services are performed". If the work is no longer expected, the capitalized amount goes to expense. Its example is an advance to a CRO.6 Companies describe it the same way in their filings: costs paid in advance are deferred as prepaid expenses and recognized as services are performed.5
The accrual entry reverses next quarter, and that is normal
Teams usually post the accrual with an automatic reversal on the first day of the next period, so the real invoice can expense normally when it lands. The net effect on next quarter's expense is the difference between what was estimated and what was billed, called the true-up. Reversal keeps the books tidy, and it also erases the evidence unless someone keeps the comparison. Keeping it is the look-back this guide returns to in section 5.
03 · What it is made of
Five cost families and the records behind them: CRO fees, site grants, labs, change orders, pass-throughs
In shortA trial's cost breaks into families, and each one needs its own way of measuring work done. Fee lines earn over time or by progress. Visits and kits earn per unit. Pass-through costs, which the CRO pays and re-bills, earn as incurred. The data lives in five or six systems that finance does not own, and Phase 3 carries the largest amounts.
Phase 3 costs more than every other development stage added together
Those are means across therapeutic areas, drawn from ClinicalTrials.gov, FDA, Medidata and IQVIA data for 2000 to 2018. The nonclinical figure is derived from a published ratio, and the authors warn that clinical costs may be understated because of reporting lags and gaps in early-phase and non-US data.10
They are figures per drug, and a drug can take several trials. Phase 3 is the late stage, which FDA describes as 300 to 3,000 volunteers with the disease or condition, over one to four years.11 A separate 2014 report for the US Department of Health and Human Services estimated cost per study at $3.80 million in Phase 1, $13.35 million in Phase 2 and $19.89 million in Phase 3, and it split that cost into components.12
No single cost family dominates Phase 3: staff, procedures and monitoring lead, each on its own clock
Share of the Phase 3 per-study subtotal, weighted averages, data from a 2014 report
Administrative staff20.4%
Clinical procedures19.8%
Site monitoring14.3%
Site retention11.5%
Nurse and clinical research associate time8.25%
Central laboratory7.46%
Physician time7.08%
Source data verification3.52%
Site recruitment3.47%
Patient recruitment2.71%
Source12 · The report's categories differ from an accrual schedule's, and it is more than a decade old. Use the spread, and take your own shares from your contracts.
Show the numbers
Item
Value
Administrative staff
20.4%
Clinical procedures
19.8%
Site monitoring
14.3%
Site retention
11.5%
Nurse and clinical research associate time
8.25%
Central laboratory
7.46%
Physician time
7.08%
Source data verification
3.52%
Site recruitment
3.47%
Patient recruitment
2.71%
Cost family
Typical lines
How work done is usually measured
Record that proves the count
Who holds it
CRO service fees
Project management, monitoring, data management
Share of term or tasks for fixed fees; visits for unit-priced lines
CRO status report, trial management system (CTMS)
CRO, clinical operations
Site costs
Investigator grants (per-visit fees paid to sites), screen failures (patients screened who do not join), start-up fees
Each completed visit at that site's budget rate
Electronic data capture (EDC) export; site budgets
Data management; CRO or sponsor site contracts
Lab and imaging
Sample kits, central reads, bioanalysis
Units processed times rate
Vendor data transfer
Central lab, imaging vendor
Change orders
Scope added by changes to the trial's written plan (protocol amendments), or re-pricing
Share done, judged with clinical operations
Signed change order; study manager's answer
Contracts, clinical operations
Pass-through costs
Patient travel, stipends, ethics fees
Cost incurred, per the report
CRO pass-through report
CRO
A common pattern we drew from practitioner guides.1,3 Your contracts decide the lines and the methods.
The sources a clinical accountant needs, and who must be asked for them
EDC is the system sites use to enter patient visit data. CTMS, the clinical trial management system, tracks site activations and monitoring visits. The CRO's accrual confirmation is the vendor's own estimate of unbilled work. Add master service agreements, work orders and change orders from the contract system, invoices and the prior accrual from the ERP, the company's finance system, and the study manager's answers to a quarterly questionnaire.
One accounting firm's guide advises a monthly sync between finance and clinical operations on sites activated, patients enrolled, visits completed and protocol changes, and tying accruals to actual activity data in place of flat monthly estimates.1
04 · The rules and their dates
The accounting and audit rules behind a clinical accrual, and what changed in 2025 and 2026
In shortNo drug regulator tells you how to compute this number. Accounting standards set when cost is recognized, audit standards from the Public Company Accounting Oversight Board (PCAOB) set how it is tested, securities rules set what happens when it is wrong, and good clinical practice governs the records it reads. As of November 2026, optional semiannual reporting is only a proposal, new expense disclosures are close, and US tax now splits domestic from foreign research.
Audit, tax and disclosure rules moved around this estimate; the way it is measured did not
Dec 15, 2020AS 2501 in force
Audit standard on estimates, for fiscal years ending on or after this date.
Jul 2025US tax law
Section 174A: domestic research deductible; foreign research stays under section 174, over 15 years.
Jul 23, 2025ICH E6(R3)
Principles and Annex 1 in effect in the EU.
Dec 15, 2025Audit tools
Technology-assisted analysis amendments, fiscal years beginning on or after this date.
May 5, 2026SEC proposal
Optional semiannual reports on Form 10-S. Proposed only.
Dec 15, 2026Expense table
ASU 2024-03: annual periods beginning after this date.
R&D cost follows the work, and non-refundable advances for future R&D are deferred, then expensed as the work is done.6,1
In force
International standards (IFRS), IAS 38
Research is expensed. Development that meets specified criteria becomes an intangible asset, so a sponsor reporting under IFRS has a further judgment on late-stage cost.20
In force; IFRS 18 changes presentation from January 1, 202719
PCAOB AS 2501
Auditors use one or a combination of three approaches: test the company's process (its methods, data and significant assumptions), develop an independent expectation, or evaluate evidence from later events. The standard also directs attention to potential management bias.13
Fiscal years ending on or after December 15, 2020; covers emerging growth companies13 (newer public companies with annual revenue under $1.235 billion, up to five fiscal years after an IPO21)
PCAOB AS 3101
Critical audit matters in the auditor's report. Not required for emerging growth companies.4
In force since the fiscal years ending on or after June 30, 2019 for large accelerated filers, an SEC filer category, and December 15, 2020 for others4
SEC Staff Accounting Bulletin 99
A 5% rule of thumb is only an initial step; small errors can be material for qualitative reasons, and errors that recur add up.22
In force
Internal control over financial reporting
Management assesses its controls each year. Issuers eligible as smaller reporting companies with revenue under $100 million do not need an auditor to attest to that assessment, and they still make it.23
A company could file semiannual reports on Form 10-S in place of quarterly reports on Form 10-Q, and the statements would still be reviewed by an independent accountant.14
Proposed May 5, 2026; comments due July 6, 2026; the SEC's file page, last updated July 31, lists no adopting release, and we found no final rule24
Accounting Standards Updates (ASU) 2024-03 and 2023-07
For public business entities, new tables split expense captions.15 Under ASU 2023-07, public companies with a single reportable segment give all the segment disclosures,25 and at least two single-segment biotechs' 2024 notes show R&D by program.26,27
ASU 2024-03: annual periods beginning after December 15, 2026, interim after December 15, 2027.15 ASU 2023-07: fiscal years beginning after December 15, 202325
US tax, section 174A
Domestic research spending is deductible, and spending attributable to foreign research stays under section 174, amortized over 15 years, so trials with sites abroad need cost split by where work happened. Ask your tax adviser how this applies.16
Public Law 119-21, July 4, 2025; amounts paid or incurred in taxable years beginning after December 31, 202416
ICH E6(R3)
Governs the trial records an accrual reads, and asks sponsors to document financial aspects and oversee transferred work (3.5, 3.6.9).9
Principles and Annex 1 in effect in the EU July 23, 2025; Annex January 2 15, 202717
Accounting rules set the amount, audit rules test the evidence, and trial rules reach only the records
Our reading of the sources6,13,4,22,16,9 · Our classification, to help a newcomer sort the rules. Section 174A governs tax, a separate calculation from the book accrual.
Show the numbers
Amount
Timing
Evidence
Disclosure
ASC 730
Governs
Governs
Touches
Touches
AS 2501
Touches
Touches
Governs
Touches
AS 3101
No reach
No reach
Shapes
Governs
SAB 99
Shapes
Touches
No reach
Shapes
ICH E6(R3)
No reach
No reach
Shapes
No reach
Section 174A
Touches
Touches
Touches
Touches
Does a change to semiannual reporting remove the accrual? No. The year-end statements would still need it and a six-month report would still be reviewed. A company electing Form 10-S would have fewer filed dates and a longer stretch between them, and any internal monthly or quarterly close would stay as it is. Plan for the rule as written today: proposed.
05 · What auditors test
What external auditors test in clinical trial accruals, and why the look-back matters most
In shortAuditors treat this as an accounting estimate built on data from outside finance. They read contracts and amendments, follow samples to invoices, ask clinical staff, and check later payments for anything left out. Comparing a past accrual with the invoices that followed is the plainest evidence you have, and teams that keep it have an easier conversation.
Auditors' own reports describe the work. In its report on MediciNova's 2024 statements, BDO USA names "Clinical Trial Accruals and Related Prepaid Expenses" as a critical audit matter, citing management judgment on "progress toward completion of specific tasks", based on data from internal clinical staff and outside service providers. Its procedures include reviewing significant agreements and amendments, evaluating trial progress documents, comparing sampled costs to invoices and contracts, checking press releases, trial databases and board materials for completeness, questioning clinical staff, and testing a sample of subsequent payments.5
One auditor's published procedures run from the contract to payments made after year end
1Read the agreements
Significant contracts and amendments.
AuditorStart
2Test the data
Trial progress documents and sampled costs, to invoices and contracts.
Our reading of the sources5,13 · Condensed from one auditor's report on one company. AS 2501 gives auditors a choice of approaches, so yours may differ.
Show the numbers
Step
What happens
Who
Time
Read the agreements
Significant contracts and amendments.
Auditor
Start
Test the data
Trial progress documents and sampled costs, to invoices and contracts.
How often does an auditor call this a critical audit matter?
No public count was found. The database that could answer it is sold by subscription. Audit Analytics put the average at 1.7 critical audit matters per US audit report in the first year of the requirement,28 which began with large accelerated filers.4
We ran a plain search of SEC filings on November 3, 2026: 49 annual reports filed in 2025 contain both "critical audit matter" and "clinical trial accruals".29 That is a ceiling, since the two phrases can sit pages apart and be unrelated. Auditors of emerging growth companies are not required to write these paragraphs.4 Read your peers' auditor reports: they tell you what your own auditor is likely to ask.
The look-back: last quarter's estimate against the invoices that followed
AS 2501 came with an amendment to the fraud standard, AS 2401, on looking back at past accounting estimates (retrospective review).13 The practical form is simple. Take the accrual at June 30, wait for the invoices covering work done by June 30, and divide. Do it by vendor and cost family, because a total that looks right can hide two errors that offset. Reviva reported this kind of work: "a detailed lookback analysis" comparing estimated accrued clinical trial expenses, specifically investigator fees from one contract research organization, with actual expenses.2
In the example, the accrual tracks the later invoices closely, and the misses run in both directions
Accrual at each quarter end against the invoices that followed for that work; the axis starts at $2 million
Accrual at quarter endInvoices that followed
$2M$2.5M$3M$3.5M
Q3 2025Q4 2025Q1 2026Q2 2026
Fictional example from this guide · Eight fictional studies. Misses that always run in one direction would be the warning sign: they suggest a bias in the method.
Show the numbers
Accrual at quarter end
Invoices that followed
Q3 2025
$2.7M
$2.6M
Q4 2025
$2.9M
$3.1M
Q1 2026
$2.6M
$2.5M
Q2 2026
$3.1M
$3.2M
One fictional study, Q2 look-back
Accrued at June 30
Invoiced later
Accrued ÷ invoiced
CRO fees
$352k
$361k
97.5%
Site costs
$196k
$204k
96.3%
Lab and imaging
$744k
$772k
96.4%
Pass-through costs
$128k
$132k
97.3%
All
$1.42M
$1.469M
96.7%
Fictional study VLT-302. Ratios by cost family use unrounded amounts. Every family came in under the invoices, which is the pattern a controller, the finance manager who reviews the entry, should ask about.
06 · The sums, worked
How to calculate a clinical trial accrual: six worked lines, from a visit count to a new change order
In shortEach method answers the same question: what has this vendor earned? Units times the rate in force on the date of the work. Contract value times the share done. Cost as incurred. A cap can lower what is earned below what the data shows. Each answer then has billed to date subtracted from it, line by line.
1 · Activity-based
Site visits, each at that site's budget
4,410 visits, priced by site = $5,402,300 earned − $5,290,100 billed = $112,200 accrual
Study NXL-301, patient visits. Pricing each visit by its own site's budget avoids an average that hides expensive sites.
Study VLT-302, monitoring. A change order raised the rate for visits on or after June 1, so May visits stay at the old rate whenever they are invoiced.
Data management, on the share of expected case report pages (the forms holding each patient's data) cleaned. The vendor had billed to 55%, through August.
Records show 251 screen failures, patients who were screened and did not join. The budget pays one for every three randomized, meaning assigned by chance to a treatment group, so 47 are not payable.
6 · A change order just started
A judgment, with the study manager's word behind it
$2,310,000 × 8% done = $184,800 accrual nothing billed until the October invoice
Signed September 12, started September 15 at 38 sites. Eight percent is an estimate from clinical operations, so the memo should say so.
All fictional, from the example in Part B.
Lab and imaging makes up almost half of this study's accrual, because one vendor stopped invoicing
$1.67MVLT-302
Lab and imaging$795K
CRO fees$382K
Change orders$185K
Site costs$178K
Pass-through costs$134K
Fictional example from this guide · Rounded, from the accrual map in Part B.
Show the numbers
Part
Value
Lab and imaging
$795K
CRO fees
$382K
Change orders
$185K
Site costs
$178K
Pass-through costs
$134K
The bridge from last quarter, the question every controller asks
A controller wants to know why the number moved. A bridge answers in five steps, from the old accrual to the new one. This one is the $112,200 site-cost line from the opening example.
Step
Amount
What it means
Q2 accrual
$76k
Estimated at June 30
Invoices against the accrual
+$6k
The true-up: invoices came in higher
Billed for Q2 work
−$82k
The old estimate is now real invoices
New Q3 work unbilled
+$112k
Done by September 30, no invoice yet
Q3 accrual
$112k
What is booked
07 · The vendor's estimate
CRO accrual confirmations: why the vendor's number and yours rarely agree
In shortA CRO accrual confirmation is a useful second opinion written by a party with its own cut-off date and its own view of the contract. Comparing it with your own build is the most informative control you have. A gap is a question to answer, and the answer belongs in the memo. Copying the vendor's number leaves you with nothing to show an auditor about your own judgment.
Sponsors commonly ask each vendor, each quarter, to state the value of work done and not yet billed. One vendor of accrual software writes that accruals built from delayed CRO invoices or month-end confirmations "are always looking backward".30 It also reports that a CRO may send a monthly summary weeks after the period ends.31 Those are a seller's words, and they match the mechanics: the confirmation is dated before the close, so anything in the last days falls out.
In the example, the sponsor's accrual runs 38% to 92% above the vendor's estimate on three explained gaps
Vendor's estimate against the sponsor's accrual, three cells
Vendor's estimateSponsor's accrual
Lab and imaging, VLT-302$434.5K$795.4K
Site costs, NXL-301$81.3K$112.2K
Pass-throughs, CVT-201$22K$42.2K
Fictional example from this guide · Fictional. Each gap has its own reason: a confirmation sent September 9 that counted only August kits; twelve patients screened after the estimate's date; a travel ceiling the estimate treated as a limit.
Show the numbers
Vendor's estimate
Sponsor's accrual
Lab and imaging, VLT-302
$434.5K
$795.4K
Site costs, NXL-301
$81.3K
$112.2K
Pass-throughs, CVT-201
$22K
$42.2K
Seven reasons the two numbers differ, and what each one calls for
Reason
What happened
What to do
Cut-off
The estimate was dated before quarter end.
Count from your own export at the cut-off, and keep your number.
Unsigned scope
Work started on a change order the vendor will not estimate yet.
Accrue what was done if it is owed, and note that the paper is unsigned.
A vendor's silence
A lab has stopped invoicing, so its billed figure stands still.
Accrue, and ask for the invoice.
Their read of the contract
The vendor treats a cap or rate differently.
Cite the clause, and write to them.
Prepaid netted away
The vendor nets unbilled work against advances.
Split by line before comparing.
Data lag
Sites have seen patients, and the records lag.
Measure the lag, and set a policy for it.
A real mistake on your side
Your model missed a site or used an old rate.
Fix the model, and keep the note.
Our reading of practitioner guides and the mechanics.1,3 No public study ranks these.
The EDC lag is measurable. A 2017 Tufts study, sponsored by a software vendor, asked data managers at 257 companies how long data takes from a patient's visit to entry in the EDC. The average estimate was 8.1 days, with wide variation (coefficient of variation .93, 230 answers).32
It is an old survey of estimates, and it gives a reason to measure your own: an eight-day lag at a September 30 cut-off means visits in the last week are thinly recorded in a first export. Ask how late your export runs, and what you do about visits done and not yet entered.
08 · Common errors
Where accrual work goes wrong: estimate errors, billing errors, and the ones nobody finds
In shortTwo kinds of error meet in this work. Estimates go wrong through missed work, stale rates and copied vendor numbers. Invoices go wrong through units above the records, rates before their date, caps ignored and duplicates. Both show up as a gap between earned and billed, so a good accrual build doubles as an invoice check. No reliable public rate exists for how often vendors over-bill.
One study's flags: the largest over-billing is $177,000 of Week 12 visits the records do not show
Items a check has marked for a person's decision (flags), on one fictional study, amount at stake
No lab invoice for August, September$730.1K
Change order started, not billed$184.8K
Visit 6 billed 300, records show 150$177K
Screen failures above the cap$30.1K
New rate applied to May 69 visits$14.5K
Fictional example from this guide · Fictional study VLT-302. The two largest are work with no invoice, to accrue; the other three are over-billing, to dispute.
Show the numbers
Item
Value
No lab invoice for August, September
$730.1K
Change order started, not billed
$184.8K
Visit 6 billed 300, records show 150
$177K
Screen failures above the cap
$30.1K
New rate applied to May 69 visits
$14.5K
Error
Why it happens
How to catch it
Cost not accrued at all
No invoice has arrived, so nothing prompts an entry. Reviva described not accruing expenses "that were incurred but for which invoices were not yet received".2
Count activity at the cut-off, for every vendor, invoice or not.
Flat monthly estimate
Easy to maintain; activity swings by site and phase.1
A change order sets a date; the invoice bills everything at the new rate.
Price each unit on its activity date.
Units above the records
The vendor bills from its own count, or from forecast.
Compare invoice units to the export, and check the lag first.
Caps and ceilings missed
They sit in an exhibit that nobody re-reads.
Carry each cap on the rate card, with its clause.
Duplicate invoices
A re-issued invoice with a suffix passes the finance system's match.
Match on vendor, period, sites and amount.
Change order late to finance
Clinical signs it, and finance hears next quarter.
Route signed change orders and started work to finance.
Spreadsheet formula breaks
One model per study, copied forward, formulas overwritten. One firm calls these methods error-prone.3
Lock formulas, review changes, or stop using one model per study.
Missed unbilled work hurts most; rate, unit and duplicate errors are cheaper, and a check can catch them
Damage if missed →
Control hard
Fix first
Note and move on
Automate it
Cost not accruedChange order lateVendor's number copiedUnits above recordsRate by invoice dateDuplicate invoice
How often it is likely to occur →
Our reading of the sources2,1,3 · Our judgment from the sources and the mechanics. No public study measures frequency. Plot your own after one look-back.
Show the numbers
Item
How often it is likely to occur (0–1)
Damage if missed (0–1)
Cost not accrued
0.42
0.78
Change order late
0.62
0.72
Vendor's number copied
0.72
0.58
Units above records
0.62
0.29
Rate by invoice date
0.34
0.20
Duplicate invoice
0.16
0.11
What no one can tell you. We looked for an independent rate of CRO billing errors and found none. The figures in circulation, such as a percentage of spend lost or a cost per trial, come from vendor blogs with no method. Treat them as unproven. Your own number is easy to get: re-check one closed quarter's invoices against the contract and the records, and count.
09 · When it goes badly
When a clinical accrual goes badly: a restatement, a material weakness and an SEC question
In shortPublic filings show the worst case. One company corrected a full year and its later interim periods after a look-back at investigator fees from one CRO showed clinical trial expenses left out of the right periods, and it reported material weaknesses. A material weakness is a serious gap in internal control over financial reporting. Below that sit true-ups that move a quarter's expense, SEC staff questions about R&D by project, and audit time. Site cash is at risk as well: sites say payment timeliness is a factor in which sponsors' studies they accept.
Reviva Pharmaceuticals' 2024 restatement, in its own filing's words
On April 12, 2024, Reviva's audit committee concluded that its 2022 annual statements and its 2023 interim statements should be restated and "should therefore no longer be relied upon", along with its Q3 2022 quarterly report. The company said it had "previously excluded certain clinical trial expenses and associated accruals from the appropriate periods".2
"Principally", it said, it had completed a detailed look-back comparing estimated accrued clinical trial expenses, specifically investigator fees from one contract research organization, with actual expenses, using historical invoices. The differences came primarily from not properly accruing clinical trial expenses that were incurred and not yet invoiced. For 2022, R&D expense, reported at $18.9 million, and related figures were "each understated by approximately $3.9 million", "representing the under accrual of clinical expenses". It expected to report material weaknesses in its 2023 annual report.2
That report's restatement note shows 2022 R&D of $18,947,874 as previously reported, an adjustment of $3,922,150 and $22,870,024 as restated, and says the company failed "to properly review and evaluate progress of expenses incurred in clinical trial contracts".7
Reviva's restatement added about a fifth to the R&D expense it first reported for 2022
Reviva Pharmaceuticals, 2022 R&D expense, as its filings state
We searched SEC 8-K filings, known as current reports, from 2019 through November 3, 2026 for those kinds of words. Of 43 hits, this was the only one filed under Item 4.02, the item for statements that can no longer be relied on.29 A word search can miss filings, so read that as "the one we found" and nothing more.
How small an accrual error can matter
SEC Staff Accounting Bulletin 99 says a 5% threshold has no basis as the only test: it can be a first step, and qualitative factors may make small misstatements material, including an error that hides a trend. It also warns that small errors can add up across years.22 For a company with little revenue, R&D is often the largest expense line, and an error in it has few other lines to hide among.
What the SEC staff asks about
One public example concerns how R&D is shown. A staff letter of June 14, 2021, quoted in the company's reply, asked Tonix Pharmaceuticals to "disclose the costs incurred during each period presented for each of your key research and development projects". Tonix's reply of June 25 described its clinical trial accrual process and how it adjusts expense when actual results differ.33
Our search of staff letters and company replies, across every year the search covers, found 13 with the exact phrase "clinical trial accruals", and this reply is the only one since 2013.29 That is a narrow test: letters can use other words. The practical lesson is that cost by project has to be available, and an accrual built by study provides it.
The site's side of a late accrual
An accrual and a payment are different things, yet the data that feeds one feeds the other. A 2017 white paper from the Society for Clinical Research Sites reports that about 40% of US sites call slow payment a primary operating concern, and that 66% of sites globally had under three months of operating cash in 2016.34
Two in three sites had under three months of operating cash in 2016
66 in 100
sites globally had under three months of operating cash in 2016, a 2017 white paper from the Society for Clinical Research Sites reports
sites globally had under three months of operating cash in 2016, a 2017 white paper from the Society for Clinical Research Sites reports
66 of 100
Research payments also feed US Open Payments reports, which ask for the recipient, the total and the study name.35 A team that can state what is owed to each site at a date has answered three questions at once.
10 · The trend
More amendments, more change orders: why the clinical accrual keeps getting harder
In shortProtocols, the written plans for trials, change more often, and each change can mean a change order and a new rate. Trials stay expensive. About 70 to 90 annual reports a year use the words "clinical trial accruals", which shows how many public filers talk about it. No public figure sizes the market for accrual software.
In Tufts' 2022 study, three in four protocols had at least one amendment, and each can mean a change order
76 in 100
protocols, Phases I to IV, had at least one amendment in Tufts' 2022 study, up from 57 in 100 in 2015
protocols, Phases I to IV, had at least one amendment in Tufts' 2022 study, up from 57 in 100 in 2015
76 of 100
The Tufts Center for the Study of Drug Development gathered data on 950 protocols and 2,188 amendments from 16 drug companies and CROs. Amendments took 260 days on average from identifying the need to the last approval, and sites ran on different protocol versions for 215 days on average. It reports 77% of amendments as deemed unavoidable.36 For an accountant, a 215-day stretch with two versions running means two prices or two visit schedules at once, depending on the site.
$40,000
a day, the estimated direct cost of running a Phase II or III trial, from 409 trial budgets.37
Tufts CSDD · 2024
$500,000
a day in lost sales from delay, in the journal article; Tufts' August 2024 white paper gives about $800,000 from the same 645-drug dataset.37,38
Two published figures
3.3
mean amendments per protocol, in the abstract's words, up from 2.1; 2,188 over 950 protocols works out to 2.3, so that figure appears to count amended protocols only.36
Tufts CSDD · 2024
About 70 to 90 annual reports a year use the words "clinical trial accruals"
10-K filings containing the exact phrase, by year filed
902021
932022
892023
792024
692025
Our reading of the sources29 · Our count in SEC full-text search on November 3, 2026. Many filers use other words, such as accrued research and development, so this understates the subject, and the drop may reflect wording or the number of listed biotechs. It measures no accrual.
Show the numbers
Period
Value
2021
90
2022
93
2023
89
2024
79
2025
69
What nobody publishes. We found no public figure for the size of the accrual software market, for the share of sponsors still on spreadsheets, or for accuracy across the industry. What you can measure instead: hours per study per close, look-back by cost family, gap to vendor estimates, and dollars disputed and recovered.
11 · AI and automation
What AI can do in a clinical accrual, what regulators say, and where plain arithmetic should stay
In shortReading contracts into rates and matching counts to contract lines are good uses of AI. The arithmetic should be ordinary code, and a person should decide judgments. The drug regulators' AI rules mostly stay out of this finance work; what binds you is internal control over financial reporting, and auditors can ask how a tool's output was checked.
Reading is where AI earns its place; the sum should stay ordinary code and the signature human
LowSomeHighEssential
Fit for AIError costNeeds a citationNeeds a personRead contracts into a rate cardEssentialHighEssentialSomeMap counts to contract linesHighHighEssentialSomeCompute earned minus billedLowEssentialSomeLowExplain a gap to a vendor estimateHighSomeEssentialHighDraft the memoHighHighEssentialHighJudge percent completeSomeEssentialHighEssentialSign the entryLowEssentialSomeEssential
Our reading of the sources13,39,18 · Our judgment. Test it on your own contracts.
Show the numbers
Fit for AI
Error cost
Needs a citation
Needs a person
Read contracts into a rate card
Essential
High
Essential
Some
Map counts to contract lines
High
High
Essential
Some
Compute earned minus billed
Low
Essential
Some
Low
Explain a gap to a vendor estimate
High
Some
Essential
High
Draft the memo
High
High
Essential
High
Judge percent complete
Some
Essential
High
Essential
Sign the entry
Low
Essential
Some
Essential
Rule or guidance
Does it reach accrual work?
FDA draft guidance on AI for regulatory decisions, January 2025
By its own words it does not address AI "used for operational efficiencies (e.g., internal workflows, resource allocation, drafting/writing a regulatory submission)" that do not affect patient safety, drug quality or the reliability of study results.40 Our reading: accrual work sits in that carve-out. It is still labeled a draft.
EU AI Act, after the AI Omnibus
The Omnibus entered into force July 27, 2026; high-risk rules apply from December 2, 2027 for listed areas and August 2, 2028 for AI in products. The areas the Commission's page names, including biometrics, employment and education, do not name accounting.41 Confirm with your counsel.
EU good manufacturing practice (GMP) Annex 11 and draft Annex 22
Those cover manufacturing. Annex 22 was still a draft on November 3, 2026.42 An accrual is finance work.
21 CFR Part 11, GAMP 5, ALCOA+
These speak to regulated records and computerized systems in regulated development and manufacturing work (GxP). EDC and CTMS exports come from such systems, and your finance copy of a count is a different thing. Let your quality team and counsel decide where the line falls, and choose controls that support your compliance.
Internal control over financial reporting
This one binds. Management assesses the controls.23 A tool that feeds the entry sits inside that assessment, and auditors can ask for evidence on its data and method under AS 2501.13
PCAOB technology-assisted analysis amendments
These address auditors' own use of technology on electronic data, in audits of fiscal years beginning on or after December 15, 2025.18 They may bring more questions about the data you hand over.
The limits to plan around when AI reads a contract
Language models read documents well and can still state something false with confidence. A 2024 academic test of commercial legal research tools found they gave false answers 17% to 33% of the time.39 That was legal research, a different task from pulling a rate out of an exhibit, and it shows why each rate needs its clause next to it. Be wary, too, of any percent-complete figure a model proposes on its own. That judgment belongs to the people who know the study.
Keep treatment groups out of finance feeds. In a blinded study, one or more parties to the trial, which can include the sponsor's own staff, do not know which treatment each patient was assigned. ICH E6(R3) asks sponsors to blind people as appropriate (Annex 1, 3.15.2), to safeguard the blinding during data entry and processing (3.16.1) and to protect it in data transfers (4.1.1).9 It names no finance feed. It follows that a kit or visit record carrying its treatment group, or arm, could show who received what, so those feeds should carry no arm.
12 · How teams do it today
How clinical accounting teams build the accrual today: questionnaires, spreadsheets and hand-offs at quarter end
In shortMost of the work is gathering. An accountant reads contracts, asks clinical operations, pulls exports, chases vendors and builds a model per study, then a controller reviews and the entry posts with a reversal. One vendor's case study puts a thorough build at a full day of several people's time per study. We found no independent figure.
From study manager to audit liaison: who touches a study's accrual
CA
Clinical accountant, the preparer
Reads contracts, builds the schedule, reconciles to vendors and drafts the entry.
Cares about: matching an invoice line to the work order that priced it.
SM
Study manager or clinical operations
Answers questionnaires on enrollment, visits and when amended work began.
Cares about: being asked once.
RC
R&D controller
Reviews the entry and memo, and asks why it moved since last quarter.
Cares about: the bridge and the evidence.
AP
Accounts payable
Posts and pays invoices, and holds what is in dispute.
Cares about: knowing what to hold, before the due date.
CR
CRO or vendor finance contact
Sends estimates and invoices, and answers variance questions.
Cares about: clear evidence, and getting paid.
AL
Audit liaison, and the external auditor
Answers samples, and tests the estimate.
Cares about: support that exists when asked for.
Eight hand-offs between a contract and a posted accrual entry
Most of a quarter-end accrual is hand-offs between finance, clinical operations and vendors
1Read contracts
Work orders and change orders into a model.
Clinical accountantBefore the close
2Ask clinical operations
Questionnaire on enrollment, visits, amendments.
Study managerClose, days 1–3
3Pull the counts
EDC, CTMS, lab and imaging exports at the cut-off.
Accountant, data managersClose, days 1–3
4Collect vendor estimates
CRO accrual confirmations, chased by email.
VendorsClose, days 1–4
5Earned minus billed
Line by line against the finance system.
Clinical accountantClose, days 3–4
6Explain the gaps
Reconcile to vendors; note the reasons.
Accountant, study managerClose, days 3–5
7Review and sign
Entry, memo, bridge.
ControllerClose, days 4–6
8Post, reverse, look back
Entry with reversal; invoices compared later.
Accounts payable, accountantNext quarter
Illustrative · replace with your own numbers · A typical pattern from practitioner guides.1,3 No public source measures these times; yours will differ.
Show the numbers
Step
What happens
Who
Time
Read contracts
Work orders and change orders into a model.
Clinical accountant
Before the close
Ask clinical operations
Questionnaire on enrollment, visits, amendments.
Study manager
Close, days 1–3
Pull the counts
EDC, CTMS, lab and imaging exports at the cut-off.
Accountant, data managers
Close, days 1–3
Collect vendor estimates
CRO accrual confirmations, chased by email.
Vendors
Close, days 1–4
Earned minus billed
Line by line against the finance system.
Clinical accountant
Close, days 3–4
Explain the gaps
Reconcile to vendors; note the reasons.
Accountant, study manager
Close, days 3–5
Review and sign
Entry, memo, bridge.
Controller
Close, days 4–6
Post, reverse, look back
Entry with reversal; invoices compared later.
Accounts payable, accountant
Next quarter
Where a study's accrual day could go: an example split of eight hours
In this example, gathering and reconciling take most of a study's accrual day
Example split of an eight-hour day per study, worked by hand
Reading contractsPulling countsEarned minus billedReconciling with vendorsInvoices, entry, memo, review
One study, one close1.5 hr1.5 hr1.5 hr1.5 hr2 hr
Illustrative · replace with your own numbers · These hours are an example. Only the full day has a published basis, and that comes from a single vendor's case study. Time your own.
Show the numbers
Reading contracts
Pulling counts
Earned minus billed
Reconciling with vendors
Invoices, entry, memo, review
One study, one close
1.5 hr
1.5 hr
1.5 hr
1.5 hr
2 hr
That case study describes one company with three to five concurrent trials, two finance staff and one reviewer: one study "consumed a full day of multiple people's time", and the close "consumed nearly a week".43 A vendor wrote it, it covers one company, and it is undated. The same vendor quotes gains of 70% to 75% in team hours and a 60% faster close.31,30 Treat all of them as claims to test on your own studies.
Three ways teams organize the work, and what each one trades
Approach
Works when
Strains when
In house, on spreadsheets
A few studies with stable contracts and an accountant who knows them well.
Study count grows, a key person leaves, or an auditor asks for a line's support two years on.
Outsourced accounting
A small company with no clinical accountant yet.
The judgment and the control stay yours, and the people doing the work are further from the study teams.
Finance system plus specialist software
Many studies and vendors. One audit firm notes that companies on large finance systems still "must go outside of those systems" to calculate accruals.3
Setup needs clean contract data, and the tool must fit your methods and your auditor's questions.
Part B · See it working
Botminds Clinical Trial Accruals on a Q3 close: eight studies, six vendors, entries due Thursday at noon
Part A described the work without naming a product. This part shows how Botminds does it, at a fictional mid-size sponsor closing the quarter that ended September 30, 2026. Every study, vendor, person, contract clause and number on the screens is sample data, and the contract text is written for the example.
13 · One study, start to finish
Tessaly's NXL-301 estimate lands at 7:55 am: from a red cell to a kept accrual and a question to the vendor
In shortThis is the example from the top of the page, inside the product. Dana Okafor, the clinical accounting lead, opens one map of every study and cost category. When Tessaly's data arrives, four AI agents build the study in view. The site-cost cell turns red with the reason already written. Dana keeps the higher number, asks Tessaly to confirm, and opens the schedule, where each amount leads to the passage behind it.
1
Eight studies by five cost categories on one accrual map
It is working day 5, Wednesday, October 7, 2026. Entries post tomorrow at noon. Seven studies were built overnight: $3.56 million of accrual against the vendors' $2.82 million.
Studies run down the side and cost categories across. Each cell holds the work done to September 30 and not yet billed, and its color says how far it sits from that vendor's own estimate: within 10%, 10% to 25%, over 25%, or missing from the estimate. A red tag marks an open billing flag. The bottom row, NXL-301, is still waiting for data.
The accrual map. The first screen of the close. Red is over 25% from the vendor's estimate; blue is work the estimate leaves out.
2
Four agents read the study while the row fills in
Dana loads Tessaly's confirmation. The contract reader turns four contracts into a rate card, the activity matcher takes the EDC, CTMS and imaging exports at the September 30 cut-off, the accrual calculator works earned against billed for five lines with 14 citations, and the billing checker runs its six checks. The panel shows each step as it runs.
Agents build the study. Contract, activity, calculation and billing checks, each with what it read and found.
3
The reason arrives with the red cell
NXL-301's site costs: $112,200 against Tessaly's $81,300, a difference of $30,900, or 38%. The panel explains it. Tessaly's confirmation, dated September 26, assumed 31 screening visits in September. The EDC cut-off shows 43. Twelve patients at the $2,575 screening grant is the whole gap.
Three source chips open the screening log, the site budgets and Tessaly's confirmation. Below sit the line, and the bridge from Q2: $76k, then +$6k, −$82k, +$112k, ending at $112k.
A cell explains itself. Our accrual, the vendor's estimate, the reason, the sources and the bridge.
4
Dana keeps $112,200 and asks Tessaly to confirm
This is her decision. She sends Tessaly the variance with its sources, and the reply will land in this cell. Keeping the higher accrual goes into the memo with its reason. The button now shows when she asked: 9:14 am.
Kept, and asked. The decision stays with the person, and goes on the record.
5
Every line of the schedule opens its evidence
NXL-301 accrues $203,800, with $20,240 in prepaid. Patient visits: 4,410 visits, each priced at its own site's budget, earned $5,402,300, billed $5,290,100, accrual $112,200. Each box in the build on the right opens the passage behind it, and the source table shows the screening note: 43 visits, 31 in Tessaly's confirmation.
The study schedule. Contract, earned, billed and accrual by line, with the build and its source alongside.
The example follows these screens exactly. The $2,575 is the sample's screening grant, and the contract terms are the sample's own text.
14 · Three readings of one map
The same accrual map read three ways: against vendors, against last quarter, and against later invoices
In shortOne grid answers three questions a controller asks. Where do we disagree with the vendor? What moved since last quarter? How good was last quarter's estimate? Open a cell or a study for the reason, the bridge and the look-back by cost category.
1
The largest gap: a lab confirmation sent on September 9
VLT-302's lab and imaging cell: $795,448 against the vendors' $434,512, 83.1% above. Larkfield's confirmation was sent September 9 and counts only 1,184 August kits ($376,512). Its September 30 transfer shows 1,112 more kits in September ($353,616), and Larkfield has not invoiced since July. Brightwater estimated 41 reads after its September 15 invoice, and the read log shows 46.
The largest gap. A stale confirmation and a vendor that stopped invoicing, in one paragraph.
2
Change since Q2 shows what moved
The second reading recolors the map by movement since the last quarter, up in blue and down in orange. Change orders new this quarter stand out, and a quiet row with a large swing is worth a question.
Change since Q2. Same cells, read by movement.
3
The Q2 look-back grades last quarter's estimate cell by cell
The third reading sets last quarter's accrual against the invoices that followed. On NXL-301's site-cost cell it reads 93.3%. Your tolerance is a setting, 5% in the sample.
The Q2 look-back. How close each cell's estimate came.
4
A study's whole bridge, with its look-back by category
Click a study and the panel shows its whole bridge. VLT-302 goes from $1.42 million to $1.68 million. The table beneath shows Q2 by cost category, 96.7% overall: invoices for work to June 30 came in $48,000 above the Q2 accrual, a true-up in Q3 expense.
A study's bridge. The answer to "why did it move?" and the look-back behind it.
Every study in the close, with its stage and preparer
Studies in the close. Contract, earned, billed, accrual, prepaid, flags, stage and preparer for each study.
15 · Invoices checked against the work
Kelbrook's invoice KC-2026-0917: five flags on VLT-302, each with the line, the clause and the count
In shortBuilding the accrual line by line also shows where invoices disagree with the contract or the records. Six checks run on every invoice. Each flag shows the invoice line, the contract clause, the recorded activity, the amount at stake and a suggested action. The preparer decides each one, and a person can add a check in plain words.
1
Five flags, $1.14 million at stake
Line 14 bills 300 Visit 6 (Week 12) visits at $1,180; the export shows 150 for the quarter, so $177,000 is billed with no recorded visit. Screen failures run over the one-in-three cap: $30,080. A new monitoring rate reaches May 69 visits: $14,490. Two more are work with no invoice: $730,128 of Larkfield kits and $184,800 on Change Order 3.
Numbered tags in each sentence open the sources. The suggested action is a proposal, and the buttons belong to the preparer.
Flags on one study. The amount, the evidence and a suggestion, and a person decides.
2
The invoice line and the visit export, side by side in the record
A numbered tag opens invoice KC-2026-0917 at line 14, highlighted: 300 at $1,180, $354,000. Another opens the EDC export at the same measure: 336 visits to June 30, 150 from July 1 to September 30, 486 to date. Only visits with a completed date count, so visits scheduled and not done stay out.
The invoice line. Line 14, highlighted in the invoice.
3
The count behind the flag
The export row is the evidence for the dispute. Before you send a letter, confirm that nothing is simply late: a sponsor can ask the study manager about visits done and not yet entered, which is why clinical operations confirms the counts before the entry goes for signature.
The visit export. Completed visits by type, at the cut-off.
4
A rate is priced on the visit's date
Change Order 2 sets $2,850 for monitoring visits on or after June 1, 2026, and keeps earlier visits at $2,640 "whatever the date they are reported or invoiced". Line 12 billed all 501 visits at the new rate, 69 of them from May 18 to 31. That is 69 × $210 = $14,490.
The change order. The clause that sets the date, highlighted.
5
A cap sits in the site budget's small print
The budget pays $640 per screen failure, "up to one screen failure for every three participants randomized across the study". With 612 randomized the cap is 204, and with this invoice 251 have been billed. All 47 on this invoice sit above it: 47 × $640 = $30,080.
The site budget. Exhibit B, with the cap highlighted.
Nine open flags across four studies, $300k to dispute and $1.01M to accrue
Flags across the close. Open flags by study and vendor, with the amount and the suggested action. Dispute letters go out with the evidence attached.
Invoice ALD-55120 was posted twice, once with a suffix: $41,225. Six expansion patients enrolled before Change Order 4 was signed: an estimated $96,000. Eighty-eight standard tumor reads on NXL-301 were billed at the expedited rate, with no written request on file: $20,240. In this sample, an amount billed above the work sits in prepaid until the preparer decides. That is the sample's policy choice; agree yours with your auditor.
On the sample close, work with no invoice outweighs over-billing, and the checks catch both
What the six checks raised, amount at stake, Q3 sample close
Work with no invoice$730K
Change order not billed$185K
Billed above recorded visits$177K
Work before a signed change order$96K
Duplicate invoice$50K
Above contract cap$30K
Wrong rate tier$20K
Above contract ceiling$20K
Billed above recorded activity$17K
Rate before its effective date$14K
Fictional example from this guide · Rounded, from the dashboard. The dashboard's $329k of billing exceptions is every billing flag added together: $300,035 open, $8,900 already disputed and $20,200 of travel the preparer accepted with a note.
Show the numbers
Item
Value
Work with no invoice
$730K
Change order not billed
$185K
Billed above recorded visits
$177K
Work before a signed change order
$96K
Duplicate invoice
$50K
Above contract cap
$30K
Wrong rate tier
$20K
Above contract ceiling
$20K
Billed above recorded activity
$17K
Rate before its effective date
$14K
16 · The entry and the signatures
CVT-201's $755,660 entry: drafted by cost family, explained in a memo, posted on two e-signatures
In shortWhen the flags are decided and clinical operations has confirmed the counts, the entry and memo are drafted. The preparer signs, the R&D controller approves with the meaning of the signature and a password, and approval posts the entry with an automatic reversal, sends dispute letters and files the audit pack. Dashboards, the forecast and the settings sit behind it.
1
A memo the auditor can read
CVT-201 is Dravenzo in pulmonary arterial hypertension, Phase 2b. The memo gives purpose, method, result, judgments, bridge and look-back in plain sentences. Echocardiography reads billed twice: disputed, $8,900. Patient travel above the $620,000 ceiling: accepted with a note, $20,200. Q2 invoices came to 96.1% of the estimate. Sam Patel has signed as preparer.
The accrual memo. Method, result, each judgment, the bridge and the look-back.
2
The controller's signature says what it means
Omar Haddad, the R&D controller, records "Approved for posting — review control performed" with his password. The signature goes on the audit trail with his name, the time and that meaning. The preparer and the approver are different people, and the entry cannot post on one signature.
Controller approval. Name, time and meaning on the audit trail.
3
Posted, balanced, reversing October 1
JE 2026-09-4484 posts to period 09-2026. It debits R&D clinical expense by cost family, and credits accrued clinical costs $755,660. It reverses automatically on October 1, and the schedule, sources, decisions, memo and signatures go to the audit pack.
The posted entry. Balanced, by cost family, with its reversal date in the header.
The close dashboard: effort, accuracy, what the checks caught, where spend is heading
The close dashboard. Accrual booked, effort per study, billing exceptions and look-back. The sample's figures are its own.
Investigator grants and CRO fees earned most of the sample quarter's $20.2M
$20.2Mearned
Investigator grants$8.6M
CRO service fees$6.1M
Central lab and imaging$3.4M
Pass-through costs$1.8M
Change orders$300K
Fictional example from this guide
Show the numbers
Part
Value
Investigator grants
$8.6M
CRO service fees
$6.1M
Central lab and imaging
$3.4M
Pass-through costs
$1.8M
Change orders
$300K
Checks, months and next quarter. Forecast against budget by study.
A study's forecast. Six months by cost family, on the same rate cards as the accrual, labeled an estimate.
A study's activity. Each agent step and each person, in order.
Lines by cost family. Includes a $3,100,000 non-refundable mobilization advance, held in prepaid until applied.
Your accrual policy, written as versioned settings
Methods and review. A method per cost family, a $250,000 controller review threshold and a 5% look-back tolerance in the sample.
Sources and people. Connected systems, and who prepares, confirms, approves and reads.
17 · Who does what
Six accrual agents, five named people, and who signs a clinical accrual entry
In shortAgents read, match, calculate, check, draft and forecast. The arithmetic is ordinary code, so the same inputs give the same answer, and each number cites its source. People decide: the preparer every flag, the study manager the counts, the controller the entry. Methods change only through an approved version, and each step goes on the audit trail. Botminds' credentials are SOC 2 Type II audit and information security managed to ISO 27001, and the controls here support your own compliance program.
Contract reader
Turns work orders, change orders, site budgets and lab agreements into a rate card: rates, fees, caps, advances and effective dates, each with its clause.
Never infers a rate that is not written. It marks anything ambiguous for a person.
Activity matcher
Maps EDC visits, CTMS monitoring visits and site-months, lab kits and imaging reads at the cut-off to contract lines, keeping the export row.
Never uses a count from outside the cut-off exports.
Accrual calculator
Works out earned against billed by the method set for each cost family, sets each cell against the vendor's estimate and explains gaps over 10%.
Never estimates arithmetic. The sums are ordinary code, with each input cited.
Billing checker
Runs six checks on invoices: billed above activity, rates before their date, caps and ceilings, duplicates, activity with no invoice for 45 days, change orders not billed.
Never decides a flag. It proposes, and shows the amount at stake.
Journal and memo writer
Drafts the entry by cost family, moves amounts billed ahead to prepaid, and writes the memo: method, judgments, decisions, bridge and look-back.
Never posts without two signatures, or writes a figure with no source line.
Spend forecaster
Projects two quarters from enrollment and visit plans, on the same rate cards, so the forecast and the accrual agree.
Never presents a forecast as anything but an estimate.
The five people on the sample's Q3 close
Fictional, from the sample data
DO
Dana Okafor · clinical accounting lead
Runs the close for eight studies and six vendors, and prepares the Veltrimab studies.
"Show me which study and which cost is off, and why."
SP
Sam Patel · clinical accountant
Prepares the Dravenzo and Talvoren studies.
"Half my week is matching an invoice line to the work order that priced it."
PR
Priya Ranvelle · study manager
Confirms enrollment, visits and amendment start dates, once per study.
"I can tell you who was screened last week. Ask me once."
OH
Omar Haddad · R&D controller
Gives the second signature on each entry and memo.
"If it moved since last quarter, I want the bridge and the evidence."
LO
Lena Ortiz · audit liaison
Exports the audit pack, and answers the auditors' samples.
"The auditors test this estimate every year. The support has to be there."
What the solution takes from your systems and what it hands back
From your contract system
Master service agreements, work orders and change orders
Site contracts and budgets
From trial systems and vendors
EDC and CTMS activity at the cut-off
Lab and imaging transfers
Accrual confirmations and invoices
Prior-quarter accruals from the ERP
Agents read and calculate, people decide and signEvery study on one map, each line tied to its clause, count and invoice
Outputs
Accrual map by study and cost category
Study schedule with evidence for each line
Bridge and look-back
E-signed journal entry and memo
Billing flags and dispute letters
Audit pack and roll-forward
What changes for the team
A gap arrives with its reason
The study manager is asked once
Support exists before the auditor asks
Connects to the records your close already runs on, by category: contract system, EDC, CTMS, vendor portal, lab and imaging transfers, and the ERP's accounts payable and general ledger. Roles separate who prepares, who confirms and who approves. Site-level activity is visible only to finance and the study team.
The design target cuts a study's quarter-end work from an example full day to 3.6 hours, most of it in the reading and pulling
Example hours per study, by hand against the design target
By hand, exampleDesign target
Reading contracts into rates1.5 hours0.3 hours
Pulling counts at the cut-off1.5 hours0.3 hours
Earned minus billed, by line1.5 hours0.4 hours
Reconciling with vendors1.5 hours1 hour
Checking invoices1 hour0.6 hours
Entry, memo and review1 hour1 hour
Illustrative · replace with your own numbers · 3.6 hours is a design target, and the split by task is an example; no independent figure exists for hours per study. The sample close's 97.1% Q2 look-back is also a design target on sample data. Measure both on your own studies.
Show the numbers
By hand, example
Design target
Reading contracts into rates
1.5 hours
0.3 hours
Pulling counts at the cut-off
1.5 hours
0.3 hours
Earned minus billed, by line
1.5 hours
0.4 hours
Reconciling with vendors
1.5 hours
1 hour
Checking invoices
1 hour
0.6 hours
Entry, memo and review
1 hour
1 hour
Part C · Ready to automate?
Should you automate your clinical trial accruals? Costs, your return, a readiness check and what to ask
Written to you, the person who answers for this number: what errors and delays cost, what automation could return, how ready you are, and how to run a measured first step.
18 · What errors and delay cost
What a weak clinical accrual costs you: restatements, audit hours, true-ups and money paid twice
In shortMost of the cost stays hidden in the close: your senior people's hours, invoices paid above the work, true-ups that surprise your CFO and audit questions answered from email. At the far end sit a restatement and a reported material weakness. Public figures are scarce, so the useful number is your own, and you can get it in a week.
$3.9M
understated 2022 R&D expense in one public restatement, against $18.9 million first reported; a look-back at investigator fees from one CRO found it.2
Form 8-K, April 2024
$0.7M
of clinical trial accruals at one company at the end of 2024, still named a critical audit matter by its auditor.5
Form 10-K for 2024
215 days
on average that sites run on different protocol versions during an amendment, each a chance of two rates at once.36
Tufts CSDD · 2024
In this example, invoices checked before payment outweigh the hours saved
An example sponsor with eight studies and $80 million a year of trial invoices, a year's value
Over-billing stopped before payment$240K
Hours saved, valued$13.4K
Illustrative · replace with your own numbers · The calculator's starting inputs in section 19, all example figures with no public benchmark behind them. Change them there: a clean vendor base shrinks the first bar, and a larger team grows the second.
Show the numbers
Item
Value
Over-billing stopped before payment
$240K
Hours saved, valued
$13.4K
Six numbers to pull from your last close
Cost
Where it hides
How to find your number in a week
Hours per study
Across accountants, study managers and reviewers, none of whom track it
Ask each person to log one study through one close.
True-ups
Next quarter's R&D expense, after the reversal
Compare last quarter's accrual with invoices for that work, by vendor and cost family.
Billing errors
Paid invoices no one compared with the records
Re-check one study's invoices for units, rates, caps and duplicates.
Missed work
Vendors with no recent invoice, change orders finance has not seen
List vendors with activity and no invoice for 45 days, and every change order signed this quarter.
Audit time
The audit fee, your staff's weeks of sample support
Count last year's requests on this area, and the days to answer them.
Vendor gap
Email threads
For one quarter, tabulate your number against each vendor's estimate by cost family.
What a clinical accrual calculator will never count
A restatement's aftermathNon-reliance filings, a reported material weakness and the remediation work that follows.2,7
Your auditor's trustAn estimate that keeps moving invites more samples, and harder ones.
Key-person riskA model only one person understands, weeks before the 10-K.
Sites paid lateIn one 2017 survey, sites said slow payment affects screening and their choice of sponsors.34
Forecasts you cannot trustIf the accrual and the forecast use different rates, the budget meeting argues about both.
19 · Your return
What accrual automation can return: preparer hours, over-billing stopped, and a look-back you can show
In shortReturn comes from three places: hours your people get back, invoices corrected before payment, and accuracy you can prove. The first two you can count; the third lowers risk. Use your own inputs, because the published figures here come from sellers and the over-billing rate has no public benchmark.
Five levers on an accrual close, each with its formula
Contract reading
Rates read once, with their clause
Change orders enter the rate card when signed, in place of next quarter's scramble.
contracts × hours saved reading
Counts
Activity pulled at the cut-off, with no chasing
Study managers confirm counts already in front of them.
studies × closes × hours chasing
Vendor gaps
Reason written with the gap
Fewer email threads, and a memo that writes down why.
gaps over 10% × hours each
Billing checks
Every invoice line against the contract
Units, rates, caps and duplicates, before accounts payable pays.
annual invoices × error share × share caught
Audit support
The pack exists before the request
Schedule, sources, decisions, memo and signatures, by study.
requests × days each × loaded cost
Your hours and dollars on your own study count
Every input marked illustrative is an assumption. Replace it with your own.
Your closeYour invoices
Illustrative yearly value
Person-hours on accruals today0
Person-hours freed0
Hours saved, valued$0
Over-billing stopped before payment$0
Yearly value, both$0
Hours freed = studies × closes × (hours today − hours with automation). Hours value = hours freed × loaded cost. Over-billing stopped = invoices × share billed above contract or work × share caught. Software and setup costs are yours to subtract.
The eight hours follow one vendor's case study of "a full day of multiple people's time",43 so your real figure may be higher. The 3.6 hours is a design target. We know of no independent public rate of CRO billing errors, so 0.5% is only an example; re-check one closed quarter to replace it. Money caught before payment is cash kept or credited; the expense for real work stays in the books.
20 · Readiness check
Is your accrual process ready to automate? Score eight questions on your contracts, data and sign-offs
In shortEight scored questions, each from 0 to 2, and a ninth for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like. Nothing you choose leaves this page.
1
Studies and vendors
How many active studies and vendors does your close cover?
Why it mattersTwo studies with one CRO can live happily on a spreadsheet. The strain arrives with more vendors and overlapping change orders.
A strong answer"Eight studies, six vendors, and more starting next year."
2
Contracts and change orders
Can you put your hands on every signed work order, change order and site budget?
Why it mattersRates, caps and effective dates live in those pages. Unsigned or missing paper is the first thing a build exposes.
A strong answer"All in one contract system, and clinical tells finance when a change order is signed."
3
Activity data at the cut-off
Can finance get EDC, CTMS, lab and imaging counts at a stated cut-off date?
Why it mattersAn activity-based accrual needs a count. If it comes through a CRO's summary weeks later, the method is only as good as the summary.
A strong answer"Scheduled exports on the first day of the close, and we know how late visits are entered."
4
Vendor estimates
Do vendors send an accrual confirmation for each study, by cost family?
Why it mattersA comparison needs a second number, on the same grouping as yours.
A strong answer"Every vendor monthly, to a template in the contract."
5
Written accrual policy
Is the method for each cost family written down and agreed with your auditor?
Why it mattersSoftware can apply a policy. It cannot pick one for you, and AS 2501 has auditors looking at whether the method fits.13
A strong answer"A policy by cost family, plus a threshold for controller review, updated yearly."
6
Look-back history
Do you compare each quarter's accrual with the invoices that followed?
Why it mattersThis is how you prove a change helped, and it is how one company found its error.2
A strong answer"Every quarter, by vendor and cost family, kept for two years."
7
Who decides and who signs
Are preparer and approver named, and is the control documented?
Why it mattersA tool can record a control. Someone still has to own it, and management has to assess it.23
A strong answer"Preparer and controller by study, with the review threshold written in the control."
8
Your auditor and IT
Have your auditor and IT seen a plan for changing how this estimate is built?
Why it mattersA change to a control mid-year is easier with no surprises. System access to EDC and the ERP takes time to arrange.
A strong answer"Our auditor saw the plan and asked for a parallel run; IT has named an owner."
9
The line an auditor picks · for reflection
If your auditor picked one accrual line tomorrow, what would have to be true for you to show the clause, the count, the invoice and the decision in under a minute?
Why it mattersYour answer lists the gaps. Each gap is a candidate job for software: reading, matching, adding up or writing down.
A strong answerA short, specific list: "every rate tied to its clause, counts kept as of the cut-off, every kept variance with a name and a reason."
0–6
Explore
Build the baseline first: time one close, do one look-back, write the method for each cost family and gather the contracts. Start with your largest study.
7–11
Ready to pilot
Run a measured pilot on one closed quarter for three studies, comparing it with what you booked and the invoices that followed. Your people decide every flag and your existing entry stands.
12–16
Ready to implement
Plan every study: contracts as rate cards, system connections, methods as settings, preparers and approvers, the memo format your auditor expects and a parallel close.
21 · Questions for any vendor
What to ask any clinical trial accrual software vendor, Botminds included
In shortAsk to see a line traced from the entry to the clause, the count and the invoice. Ask how arithmetic is done, how methods change, who can see site-level data and what you can give your auditor. Ask us too. A good answer is shown on your own documents.
A
Reading contracts and citing the clause
A1
Show me a rate on the card. Where is its clause and page?
Why it mattersA language model can be confidently wrong.39
Strong answer · warning sign
One click to the passage, on my contract.
"It's accurate. Trust the rate card."
A2
What happens when a clause is ambiguous, or a change order is unsigned?
Why it mattersGuessing a rate here becomes a wrong number downstream.
Strong answer · warning sign
It marks it for a person, and the memo records the decision.
It picks the likeliest value.
B
Methods, arithmetic and the vendor gap
B1
Is the arithmetic ordinary code or a model's output?
Why it mattersThe same inputs must give the same answer every time.
Strong answer · warning sign
Deterministic, with each input shown.
"Our AI calculates the accrual."
B2
Can each cost family use its own method, and each rate apply by activity date?
Why it mattersA change order rate and a percent-complete fee work differently.
Strong answer · warning sign
Methods are settings, versioned and approved.
One method across the study.
B3
How do you deal with visits done and not yet entered at the cut-off?
Why it mattersEntry lag is real, and a 2017 survey put it at 8.1 days on average.32
Strong answer · warning sign
A stated policy, and the study manager confirms.
"We just use the export."
C
Your controls and your auditor
C1
What will you give my auditor, and has an auditor accepted it elsewhere?
Why it mattersAS 2501 has them testing methods, data and assumptions.13
Strong answer · warning sign
A pack per study; references on request.
"Auditors love it."
C2
Which assurance reports do you hold, and do they cover what my financial reporting relies on? Would you need a SOC 1?
Why it mattersYour auditor may ask about a service provider's controls over financial reporting. Check with them early, since reports differ in purpose.
Strong answer · warning sign
Exact report names, periods and scope, stated plainly. Botminds: SOC 2 Type II, information security managed to ISO 27001.
"We're SOX compliant." No product can be.
C3
Who can post? Can one person prepare and approve?
Why it mattersSeparation of duties is a basic control.
Strong answer · warning sign
Two named signatures, each with its meaning, on an audit trail.
An admin can override silently.
Deeper questions for a working session: data, blinding, change and exit
D
Trial data and who sees it
D1
Does any finance feed carry a treatment arm?
Why it mattersICH E6(R3) asks sponsors to safeguard the blinding, including in data transfers (3.16.1, 4.1.1).9 A record with an arm on it could show who received what.
Strong answer · warning sign
No arm in finance feeds, with who can see what written down.
"We haven't thought about that."
D2
Who can see site-level activity and payments?
Why it mattersSite budgets are commercially sensitive.
Strong answer · warning sign
Finance and the study team, by role.
Everyone with a login.
E
Change, scale and exit
E1
How do you add a check or a new vendor's format?
Why it mattersEach CRO's report looks different.
Strong answer · warning sign
Your team can describe a check, test it on this close and approve it.
A paid change request each time.
E2
Can I export schedules, decisions and memos if I leave?
Why it mattersAuditors can ask about past years.
Strong answer · warning sign
A plain export of every record and citation.
Reports only, while you subscribe.
E3
How do reports change if we move to semiannual filing or IFRS 18?
Why it mattersThe first is a proposal and the second has a date.14,19
Strong answer · warning sign
Periods and groupings are settings.
"We'd build that."
22 · How to start
A twelve-week accrual pilot: one closed quarter, three studies, measured against what the invoices said
In shortPick a quarter that has already closed, so the invoices that followed are the answer key. Rebuild three studies, compare with what you booked and with the invoices, then run the next live close in parallel. Keep your existing entry until you have the numbers.
Prove it on a closed quarter, then run one live close in parallel, before anyone changes a control
Your teamVendor
Baseline: time, look-back, gaps
Load contracts for three studies
Connect exports and invoices
Rebuild the closed quarter
Compare with booked and invoices
Walk your auditor through it
Parallel run on the live close
Measure and decide
Week 036912
Illustrative · replace with your own numbers · Twelve weeks is an assumption. Line it up so the parallel run meets a real close.
Show the numbers
Activity
From
To
Baseline: time, look-back, gaps
0
2
Load contracts for three studies
1
4
Connect exports and invoices
1
4
Rebuild the closed quarter
4
6
Compare with booked and invoices
5
7
Walk your auditor through it
7
8
Parallel run on the live close
8
11
Measure and decide
11
12
Choose three studies that disagree
One large, one with change orders, one thin
A Phase 3 with many sites, a study mid-amendment, and an early study with a milestone contract. Three kinds of method, and the disagreements you already know about.
Plant a few errors in copies
Test what the checks catch
In copies of invoices, add a duplicate, a rate before its date and units over a cap. Count what is found, what is missed and what is wrongly flagged.
Seven measures for the pilot's scorecard
Measure
How
A pass looks like
Person-hours per study
Logged in the baseline and in the pilot
Your target, set before you start
Look-back by cost family
Pilot accrual against the invoices that followed
Inside your tolerance, and no worse than what you booked
Rate card accuracy
Reviewer checks each rate against its clause
No wrong rate passes unmarked
Gaps explained
Share of gaps over 10% with a correct, cited reason
Your team agrees with the reason
Billing catches
Planted and real errors found
All planted errors, few false flags
Trace time
Stopwatch, entry to clause, count and invoice
Under a minute
Your auditor's view
A walk-through of one study's pack
Their questions are listed and answered
No public benchmark exists for most of these. Your baseline is the benchmark.
Start with the cost families that have clean counts and large gaps
Size of the gaps →
Fix the data first
Start here
Leave for later
Quick win
Lab and imagingSite visitsChange ordersCRO fixed feesPass-throughs
How clean the count data is →
Illustrative · replace with your own numbers · Our judgment for a typical sponsor. Plot your own after one look-back.
Show the numbers
Item
How clean the count data is (0–1)
Size of the gaps (0–1)
Lab and imaging
0.72
0.74
Site visits
0.58
0.62
Change orders
0.26
0.68
CRO fixed fees
0.34
0.30
Pass-throughs
0.66
0.22
23 · FAQ
Clinical trial accruals: the questions accountants and controllers search for
In shortEach answer stands alone, in plain words, with the detail in the sections above.
What is a clinical trial accrual?
A clinical trial accrual is the expense a sponsor books at a period end for trial work that has been done and not yet billed. It covers CRO fees, site payments, labs, imaging and pass-through costs. The amount is usually worked out as earned to date minus billed to date, line by line.
How do you calculate clinical trial accruals?
For each contract line, work out what the vendor has earned: units times the rate in force on the activity date, a fixed fee times the share done, or cost as incurred. Subtract what has been invoiced. A positive result is an accrual and a negative one is a prepaid, and caps or ceilings can lower what is earned.
What is a CRO accrual confirmation?
It is a contract research organization's own estimate of the work done and not yet billed on a study, usually sent monthly or quarterly. It is useful as a comparison and is dated before the period ends, so it can miss late activity. A company should still form its own view for accounting and audit.
What is the journal entry for a clinical trial accrual?
Debit R&D expense and credit accrued liabilities for work done and not yet billed. If more has been billed than earned, the difference sits in prepaid, an asset, until the work happens. Teams usually post the accrual with an automatic reversal on the first day of the next period, so the real invoice can expense normally when it lands.
How do you reconcile CRO invoices to the work done?
Compare each invoice line with the contract and the activity records: units against the visit or kit export, the rate against the one in force on the activity date, amounts against any cap or ceiling, and the invoice against others for duplicates. Check for data-entry lag before you dispute a count. A person should decide each difference and keep the reason.
How are non-refundable advances to a CRO accounted for?
Under US GAAP, ASC 730-20 has non-refundable advance payments for future R&D goods or services deferred and capitalized, then expensed as the goods are delivered or the services performed. In practice the advance sits in prepaid and turns into expense as the work happens.
Why do auditors focus on clinical trial accruals?
The accrual is an estimate based on data from outside finance, and it can be large relative to everything else on a clinical-stage company's statements. One auditor's published critical audit matter cites the judgment needed to estimate progress, using data from clinical staff and outside providers. Its procedures reach contracts, invoices and payments made after year end.
Are clinical trial accruals a critical audit matter?
Sometimes. The label depends on the auditor's judgment for that company and year, and critical audit matters are not required for emerging growth companies. No public count was found. One company's 2024 auditor report used it for a balance of $0.7 million of accruals.
What is a look-back in clinical accruals?
A look-back compares a past period's accrual with the invoices that later arrived for that work. Running it by vendor and cost family shows whether estimates run high or low. It is also strong evidence for an auditor, and one company reported such an analysis in explaining its restatement.
What happens if clinical trial accruals are wrong?
A small miss becomes a true-up in the next period's expense. A large one can mean correcting past statements. In 2024 one public company said its 2022 R&D expense was understated by about $3.9 million, because clinical trial expenses were left out of the right periods, as a look-back at investigator fees from one CRO showed, and it expected to report material weaknesses.
Which method should we use for CRO fees and site costs?
Match the method to how the work happens and how the contract pays. Teams often use percent complete or elapsed time for fixed fees, units for monitoring, kits and reads, each visit at its site's budget for grants, and cost incurred for pass-throughs. Write the method down by cost family and agree it with your auditor.
Does semiannual reporting remove the need for quarterly accruals?
Not as of November 2026. Optional semiannual reporting on Form 10-S is an SEC proposal from May 5, 2026, with no adopting release found. Even if adopted, annual statements and six-month reports reviewed by an accountant would still need the accrual, and many companies would keep an internal close.
What is clinical trial accrual software?
It is software that reads or stores trial contracts, takes activity counts at a cut-off, works out earned against billed and supports the entry and memo. Tools differ in how much they read for you and how well each number traces to its source. Test any tool on a closed quarter, against the invoices that followed.
Can AI build a clinical trial accrual?
AI can read contracts into rates, map counts to contract lines and draft a memo, with each value cited. Arithmetic should stay ordinary code, and people should decide judgments such as percent complete and sign the entry. FDA's January 2025 draft guidance excludes operational uses that do not affect patient safety, drug quality or study results; internal control over financial reporting still applies.
24 · One-page summary
Clinical trial accruals on one page: the method, the 2026 rules, readiness and 14 terms
Print this page
In shortThe whole guide on one screen: the method, the rules as of November 2026, the readiness bands and the terms.
The method
Read every contract and change order into rates, caps and effective dates, with the clause.
Set a cut-off, and take counts from EDC, CTMS, lab and imaging exports.
For each line, work out earned, using the method for its cost family.
Subtract billed from the ERP: an accrual if positive, a prepaid if negative.
Compare with each vendor's estimate, and write down the reason for each gap.
Check invoices for units, rates, caps and duplicates. A person decides each flag.
Study manager confirms the counts; preparer and controller sign; the entry posts and reverses.
Next quarter, compare with the invoices that followed.
Readiness: score 0–2 on eight questions: 0–6 explore · 7–11 pilot · 12–16 implement.
As of November 2026: SEC semiannual reporting proposed only · ASU 2024-03 annual periods beginning after December 15, 2026 · IFRS 18 from January 1, 2027 · section 174A splits domestic from foreign research · E6(R3) Annex 2 in the EU from January 15, 2027 · FDA's January 2025 AI draft sets operational uses aside.
14 terms
Accrual
Work done and not yet billed, booked as expense and a liability.
Prepaid
Billed or paid ahead of the work; an asset until earned.
Earned to date
The value of work done so far under a contract.
CRO
Contract research organization, a firm hired to run trial work.
Accrual confirmation
A vendor's own estimate of its unbilled work.
EDC
Electronic data capture: where sites enter visit data.
CTMS
Clinical trial management system: site and monitoring records.
Change order
A signed change to a work order's scope or price.
Pass-through
A cost the CRO pays and re-bills at cost.
Screen failure
A patient screened who does not join the trial.
True-up
The later correction when real invoices differ from an estimate.
Look-back
A past accrual compared with the invoices that followed.
Critical audit matter
A matter an auditor describes in its report as especially challenging.
Material weakness
A serious gap in internal control over financial reporting.
25 · Sources
Sources
Accessed November 2026. Regulatory status statements are as of November 2026.
D. Brownstein, Haskell & White, "Tackling Clinical Trial Accruals: A Guide for Med Tech Finance Leaders", May 6, 2025. hwcpa.com
Reviva Pharmaceuticals Holdings, Form 8-K, Item 4.02, filed April 15, 2024. sec.gov
BDO, "Automating and Simplifying Accrual Management for Clinical Stage Biotech Companies", August 9, 2024. bdo.com
SEC, Order approving the Auditor Reporting Standard (AS 3101), Release 34-81916, October 23, 2017. sec.gov
MediciNova, Form 10-K for 2024 (auditor's report February 19, 2025). sec.gov
Deloitte, Life Sciences Industry Accounting Guide, Research and Development, March 2024, section 3.2.4, summarizing and quoting ASC 730-20. deloitte.com
Reviva Pharmaceuticals Holdings, Form 10-K for 2023, including Note 2 (restatement). sec.gov · Note 2
Sertkaya, Beleche, Jessup, Sommers, "Costs of Drug Development and Research and Development Intensity in the US, 2000-2018", JAMA Netw Open 2024;7(6):e2415445. jamanetwork.com
FDA, "Step 3: Clinical Research", Drug Development Process. fda.gov
Sertkaya, Wong, Jessup, Beleche (ERG for HHS ASPE), Examination of Clinical Trial Costs and Barriers, 2014. aspe.hhs.gov
SEC, Order Granting Approval of AS 2501, Auditing Accounting Estimates, Release 34-86269, July 1, 2019, and SEC notice of filing, Release 34-85434, 84 FR 13396, April 4, 2019. sec.gov · govinfo.gov
SEC, Semiannual Reporting, Federal Register Vol. 91 No. 88, May 7, 2026. govinfo.gov
FASB, Accounting Standards Update 2024-03, November 2024, and ASU 2025-01, January 2025 (effective date). ASU 2024-03 · ASU 2025-01
Draft EU GMP Annex 22 (consultation July 7 to October 7, 2025). picscheme.org
Condor Software (an accrual software vendor), "How a lean biotech finance team scaled trial volume without adding headcount", case study, undated. condorsoftware.com
Next step
See it on your own studies.
Explore the solution, or bring one past quarter of your contracts, CRO estimates and invoices, and we will build the accrual with you, every line tied to its clause, its count and its invoice.
We use essential cookies to make this site work, and cookieless, anonymised analytics to understand how it's used. "Accept all" also turns on Google Analytics, which sets cookies. Privacy policy