GCP inspection readiness, explained.

What a GCP inspection is: a regulator checks the story of a trial against its records

One inspection, from the file a trial leaves behind to the regulator's classification

1. The trial runs and leaves its recordNorthwind Bio's study VLT-302 runs at 46 sites. Monitors visit, departures from the study plan (deviations) are logged, and documents reach the eTMF, some of them late.
2. The notice arrivesFDA calls on October 5 and names October 26 to 29. The clock on all of the preparation starts here.
3. The sponsor gets readyTopics are chosen, storyboards written and checked against the record, gaps given owners, people rehearsed, and a working room, the back room, set up to find documents. This is where the training sentence is caught.
4. The inspection weekFDA's inspectors interview people in a meeting room, the front room, and ask for documents. Requests go to the back room, and every document sent is logged.
5. Findings, response and classificationFDA lists any observations on a Form 483. The sponsor may respond within 15 business days. FDA classifies the inspection and posts the result.

Step 3 is the subject of this guide. The people, companies and figures in this example are fictional.

Why a slow document search counts against a sponsor, and three more surprises

Storyboard, front room, back room, note-to-file: the words this work runs on

Fourteen terms are defined in the one-page summary.

FDA BIMO, the MHRA and EMA: who inspects a sponsor, why, and how much warning they give

In shortFDA usually inspects in support of an application, and its general minimum warning is five calendar days. The MHRA asks for a dossier within 30 days and then agrees a date. At the European Medicines Agency (EMA), a committee requests inspections for centralized applications, those covering the whole EU, and national inspectorates carry them out. FDA and EMA inspect far more investigator sites than sponsors; in EMA's 2024 count, sponsors drew the most findings per visit.

FDA closes an inspection with one of three labels: No Action Indicated (NAI), Voluntary Action Indicated (VAI) or Official Action Indicated (OAI). One word needs care from here on. FDA calls its inspectors investigators, and the doctor who runs a trial at a site is a clinical investigator.

FDA also says it uses the same approach for foreign and domestic inspections in most respects. On foreign data, the September 2026 message from its center directors is direct. Companies should not assume that studying abroad lowers the chance of inspection. FDA is adding resources for foreign inspections, and it is looking harder at foreign phase 1 and other early-stage studies.2,12

How FDA chooses which sites to inspect, and when it adds the sponsor

FDA's December 2024 training slides say GCP inspections "are not required" in application review. Inspections of clinical investigators are usually requested for new molecular entities, active ingredients FDA has not approved before, and for efficacy supplements, changes to an approved drug, that add a dose, a population or an indication.

On average, two or three sites per protocol are chosen, with help from a scoring tool whose inputs include enrollment, efficacy outcome, complaints, protocol deviations and the time since the last inspection. Depending on what those inspections find, FDA may add a sponsor or CRO inspection or send information requests.13

The agency's public page on risk adds a site's compliance record and data irregularities to the list.14 We could not confirm the longer lists of sponsor triggers that circulate in trade writing, so treat any such list as someone's experience.

Clinical investigators are the large group: 609 final classifications in fiscal 2024, against 99 for sponsors and CROs and 10 for sponsor-investigators.15 Across every BIMO program, a law firm's reading of FDA's figures has about 650 inspections in fiscal 2021 and over 1,000 in fiscal 2024 and 2025 to date.17 That total includes laboratories, ethics boards and other programs outside this guide. FDA also runs remote regulatory assessments, which are not inspections; it completed none for sponsors and CROs in fiscal 2024.15

In the EU, EMA reports GCP inspections up 24% in 2024 and 26% in 2025, gives no counts on those pages, and credits capacity building for the 2025 rise.18,19 This guide follows FDA, the MHRA and EMA; Japan's PMDA, Health Canada, Swissmedic and Australia's TGA run GCP inspections of their own.

The GCP rules behind an inspection, and what changed between 2021 and 2026

In shortICH E6(R3), the latest revision of the good clinical practice guideline from the International Council for Harmonisation, is the new shared text. It asks sponsors for oversight in proportion to risk, and for records that are timely and directly accessible. A 2022 law spelled out FDA's BIMO inspection authority, and FDA published its inspection practices in December 2025. The UK's amended clinical trial regulations took full effect on April 28, 2026.

In the table, 21 CFR is title 21 of the US Code of Federal Regulations, which holds FDA's rules.

Nine GCP inspection areas, and the systems that hold the answer to each

In shortFDA's sponsor program has many sections, among them organization, investigators, CROs, monitoring, data, electronic records, records custody and the investigational product. Many sponsors group this work into areas. The answer to any one question usually sits in two or three systems, and that is the reason a by-hand search takes so long.

Nothing says a sponsor must use nine areas. We use the nine in the fictional example because they follow much of how FDA's program divides the sponsor's duties. FDA's document says it provides "only the minimum scope", and that investigators should go further if they find a problem touching participants or data reliability.1

The program also has sections beyond these nine, including quality assurance, ClinicalTrials.gov and financial disclosure, and this guide, which is about trial conduct, does not go through all of them. A real plan covers every section that applies to your studies.

Terms in the table: delegation is a site's record of which staff member may do which task. Blinded people do not know which participant received which treatment, and unmasked data would show them. A query is a question sent to a site about a data point. Database lock is the point when a trial's data are frozen for analysis.

What an inspection storyboard holds, and the test every sentence has to pass

In shortA storyboard tells one topic in a fixed order, in short factual sentences, each one traceable to a record. Regulators do not ask for it and no template is official. The test is plain: can you put your finger on the document that proves this sentence? The answer is yes, partly, or no.

The two published descriptions we could find are practitioners' columns. One describes a one- to two-page document that refers to TMF records and procedures without copying them, with dated events in order, short statements of action and a brief statement of regulatory risk.3

The other says there is "no standard template", that PowerPoint, Word and Excel all serve, and that a storyboard can grow out of the risks a team identifies during a trial or an issue when it appears, with updates as the response progresses.4 A consultancy's checklist lists "documentation, storyboards, and presentations for known gaps" under operational readiness.33 A common shape has eight parts.

Frame 4 names two groups a study's statisticians analyze. The full analysis set stays as close as possible to everyone randomized. The per-protocol set is the part of it that kept more closely to the study plan.

Supported, partly supported, not in the record: reading a storyboard sentence by sentence

Take the sentence from the example at the top of this page: "The new coordinator, who started on March 2, was trained on the visit schedule before she booked her first visit." The log shows protocol training on March 3, data system training on March 4, and visit-schedule training only on March 20. She booked a Week 16 visit on March 9. So the sentence is not in the record.

Now take "our monitors check visit windows at every visit." Monitoring plan version 3 says they check at every on-site visit, and remote visits are a different matter. That sentence is partly supported. A sentence that fails has three honest remedies.

• Find the missing record. If a training record from March 5 exists somewhere, file it honestly and the sentence stands.
• Reword to what the record says. "She was retrained with the whole site team on March 20." Less comfortable, and true.
• Raise it as a finding of your own. A real gap in training is itself a fact, and it reads better coming from the sponsor first.

How the front room and back room work during a sponsor inspection, and why retrieval time matters

In shortThe front room answers questions; the back room finds, checks and logs documents. FDA says records should come "in a timely manner" and names no number, and we found no public benchmark for retrieval time. The one published warning is the MHRA finding where documents took two days. Measure your own time in a rehearsal.

FDA's guidance says the sponsor may ask clarifying questions about a request, "should be prepared to provide the records in a timely manner", and should be ready to give access to electronic systems on arrival.2 A consultancy's checklist adds the working parts: a "backroom hub", QA document review, runners and scribes.33 An ACRP article recommends logging every requested document, making two copies, answering only what is asked, and holding a daily wrap-up.34 Keeping the second copy means you know exactly what the inspector holds.

A median tells you more than an average here, because one large request can distort the mean. In a fictional rehearsal with four requests, taking 0:22, 0:38, 0:44 and 1:05 (minutes and seconds), the median is halfway between the two middle values, 0:38 and 0:44, so 0:41. The single slow one, six sets of oversight minutes, is worth a look of its own: which system held them, and who had to be asked?

Documents filed after the inspection notice: what the audit trail shows and how a note-to-file helps

In shortFile what is missing, with today's true date, and say why it was late. Never backdate. A note-to-file explains a late or odd record, and a note cannot stand in for a record that never existed. No regulator publishes a rule on how many late filings are too many; they read the pattern.

E6(R3) says essential records should be "collected and filed in a timely manner" and that alterations "should be traceable", without naming a number of days.20 EMA tells sponsors to define filing timelines in their procedures or TMF plan, so a figure such as 30 days is a sponsor's own target.31 Source records should be "attributable, legible, contemporaneous, original, accurate and complete", the set often called ALCOA+; the MHRA says the plus adds complete, consistent, enduring and available, with "no difference in expectations".20,35

The week after a notice is when this goes wrong. Someone finds a folder of site documents that were never uploaded, and the team files them all. That is the right thing to do, as long as the file then shows the truth. Part 11 requires secure, time-stamped audit trails, and FDA's investigators can read them, so a burst of filings on October 6 for documents dated in the spring is plainly visible.27,2 The MHRA critical finding we quoted earlier cited that pattern, alongside documents not uploaded in a timely manner.6

What belongs in a note-to-file, and four ways a note can hurt you

We found no regulator's template. The MHRA's guidance on inspection responses does say an organization should decide whether evidence can be corrected "or whether the problem requires documentation only (e.g. in a file note, deviation record etc)".11 Working from the principles above, a useful note gives the document and its date, the date filed, who filed it, the factual reason it was late, and a signature from the person responsible.

Published GCP inspection findings: essential documents, source records, monitoring and data management lead

In shortFDA, EMA and the MHRA all publish what they find, in different forms. Among the critical and major findings EMA reports for 2024, essential documents, source documentation and monitoring lead. FDA's sponsor list starts with monitoring and following the plan. These are the subjects where the story and the record part ways.

The FDA and EMA totals here count findings on every subject, and this section describes those on trial conduct. FDA's two decks also count differently: 125 is the sponsor inspections in its fiscal 2024 Form 483 trends deck, and the 99 in sections 02 and 09 is sponsor and CRO inspections given a final classification that year.15,36

Sponsors stand out when you divide by visits. EMA's committee sent inspectors to 18 sponsors and 42 investigator sites in 2024. Sponsors drew 152 findings and investigators 165.8 A sponsor inspection covers more of the trial, so this says little about whether sponsors are worse. It does tell a sponsor what to expect.

A selection of the trial-conduct findings regulators describe

What a bad sponsor inspection can cost: from a Form 483 to rejected data and a delayed approval

In shortMost FDA sponsor inspections end with no action. The bad outcomes are rare and heavy: rejected data, a clinical hold, an application held up. The MHRA grades every finding, and every commercial sponsor in four years of its reports had at least one major or critical finding. FDA does not publish how often an inspection delays an approval.

FDA uses "official action" when the scope, severity or pattern of violations show that participants were exposed to unreasonable risk, that their rights or welfare were seriously compromised, or that "data integrity or reliability is or has been compromised". The tools then include an untitled or warning letter, rejection of data, seizure, injunction, civil money penalties, prosecution under the Federal Food, Drug, and Cosmetic Act (FD&C Act) or other federal statutes, and a clinical hold, an FDA order that delays or suspends a study.1

Section 704(a)(5) of that law says that sponsors must give FDA access to records and systems, and FDA now tells sponsors that being unable to inspect or to reach data is "a material factor" in regulatory strategy.2,12 FDA posts its final classification publicly; a law firm reading suggests FDA may also post more BIMO Form 483s.2,39

At the MHRA, a critical finding goes to a cross-agency Inspection Action Group, which may require periodic reports, return early, refer the case to other bodies, suspend trial authorizations, issue an infringement notice or prosecute.7 A major finding left unaddressed from the last inspection can be raised to critical on the next.11 The TMF has its own line: one "incomplete to such an extent that it cannot form the basis of inspection" can be critical by itself.7 The MHRA's 2018 to 2019 report shows one case of that escalation.6

How much a day of delay can mean, worked three ways

There is no public figure for how many approvals an inspection has delayed. If one delayed yours, the Tufts Center for the Study of Drug Development gives two ways to price a day, from one dataset of 645 drugs. Direct Phase II and III trial cost is about $40,000 a day. Lost sales are about $500,000 a day in the journal article and about $800,000 in the August 2024 white paper.41,42

Those are estimates across many drugs, and a lost-sales day exists only if your launch date moved. The chart runs a 30-day delay, an assumption, through each.

How to put a number on inspection readiness, and what that number can and cannot tell you

In shortNo regulator uses a readiness score. It is a management tool, and it is worth having if you can open it and see what it rests on. Count sentences backed by the record, open gaps by severity and by due date, questions rehearsed, and back-room time. Report them by area so the weakest one shows.

Start with the counts, since they can be checked. In the fictional study, 274 of 291 sentences are backed: 274 ÷ 291 = 94.2%. That leaves 11 partly supported and 6 not in the record. Then ask how much weight each deserves. One unsupported sentence about training at one site is a smaller risk than an unsigned memo behind 214 late data changes. That is why gaps carry a severity, and why a good team watches the four high-severity gaps before it watches the percentage.

The two sponsor topics that get the hardest questions: oversight of the CRO and data changed before lock

In shortA sponsor can transfer the work and keeps the responsibility, so an inspector will ask how it knew the CRO was doing the work well. And any change to data near database lock has to show who approved it and why, in a form the audit trail confirms. In both, "we always do that" is the sentence that fails.

What a sponsor keeps after transferring obligations to a CRO

Under 21 CFR 312.52 a sponsor may transfer any or all obligations to a contract research organization (CRO), but the transfer "shall be described in writing", and "any obligation not covered by the written description shall be deemed not to have been transferred".24

FDA's program tells investigators to obtain the contract and any separate transfer document, check that the sponsor told FDA of the transfer, and document any that went unreported. Device sponsors are different: Part 812, FDA's rule for device studies, does not provide for transfer, so they answer for a CRO's noncompliance.1 E6(R3) says responsibility for trial quality and data integrity stays with the sponsor, that activities not specifically transferred are retained, and that oversight should be "fit for purpose".20

Consider two gaps from the fictional study. Monitoring and data management both went to a CRO, while the form sent to FDA lists monitoring only. And the signed monthly oversight minutes stop at June, with July and August missing from the TMF. A front-room answer that says "those meetings carried on every month" has no document behind it for those two months. Two honest ways out exist: find and file the minutes, or say the record stops at June and explain.

Who approved 214 changes after soft lock, and how a record answers

A soft lock is when a database is frozen for final review, and only approved corrections may enter. A hard lock ends even those. FDA's program asks about access to unmasked data before lock, and tells investigators to read how queries reach sites and whether monitoring verified the corrections.1 E6(R3) says audit trails should stay switched on and be interpretable. EMA expects "risk-based trial specific audit trail reviews", with review documented and focused on critical data.20,32

Those are the counts a person needs in the room. 176 ÷ 214 = 82.2% approved, and 15 ÷ 22 = 68.2% of the medication dates confirmed. A sentence that gives the true number and what is being done about the rest stands up better than one that rounds up to "all".

Mock inspections and inspector questions: rehearsing people without coaching them past the record

In shortA mock inspection tests people and process under pressure. Coaching means teaching someone to answer the question asked, with the exact number, and to say "I will check" when unsure. It never means teaching a better story than the record holds. Federal law applies to what is said to an FDA investigator.

A hired mock inspection usually lasts a fixed three to five days and draws on FDA's compliance programs. One consultancy also describes a one-day exercise aimed at the inspection process and back-room operations, with a QA facilitator in each interview.33 A real FDA inspection has no fixed length.33,2 A rehearsal is worth more when the person answers in their own words and someone holds that answer against the file. It is worth less as a script, since a person who memorized one has no way to answer the follow-up.

The legal line behind "if you do not know, say you will check"

FDA's program lists prosecution under federal statutes, naming 18 U.S.C. 1001 among them, and tells investigators to raise potential fraud with supervisors.1 Section 1001 covers false statements in matters before the federal government. We are not giving legal advice here, and your counsel should brief anyone who will be interviewed.

The practical rule is short. A person who says "I do not know, I will check" has done nothing wrong; a person who guesses at a date has made a statement to a federal investigator. The MHRA's 2017 to 2018 report shows a regulator grading false and misleading information as a critical finding in its own right, separately from whatever the underlying problem was.38

What AI changes in GCP inspection readiness, what regulators say about it and what stays with people

In shortAI is good at the slow middle: reading many records, drafting from them, and comparing a sentence with a source. Regulators have said little about this use, and we found no rule on AI-drafted storyboards. The defensible design is that software proposes and cites, and a named person signs, answers and releases.

A language model that drafts a storyboard can write a sentence that sounds right and is not in the record: the same failure this whole practice exists to catch. So the control belongs in the design of the tool. Every sentence must point to a passage. A person or a second check must be able to open that passage. A sentence that cannot be traced gets flagged, however good it sounds.

The test on any tool is what happens to the sentence with no source. People keep the decisions, the signatures and the answers in the room, because the inspector is talking to them.

Any tool that signs storyboards electronically also falls under Part 11 and your own computer system validation. No product can be certified for GxP, the good-practice rules for regulated work, or for Part 11; a vendor can offer controls that support your compliance, and validating your use remains your job.27,29 Plan to tell an inspector, if asked, how a document was prepared and who checked it. The EMA paper's phrase about model details being requested at inspection is a useful planning assumption for any use nearer the data itself.44

How clinical QA teams prepare for inspection today: the roles, the hand-offs and where the days go

In shortClinical QA leads, and five other roles carry topics, documents and answers. Storyboards are built by hand from several systems in slides and spreadsheets. Most of the time goes to finding documents and checking sentences. Nobody publishes how long this takes or what it costs, so measure one storyboard of your own.

Who holds what in a sponsor's inspection team

Three ways sponsors staff inspection preparation today

We found no public figure for what sponsors spend on inspection preparation, how long a storyboard takes, the size of this market, or how fast a back room should work. The estimate above is exactly that. Part C lists what to time in a week.

October 7 at Northwind Bio: from the failing storyboard frame to a signature the record supports

In shortDana Okafor, the clinical QA lead, opens the war room 19 days before FDA arrives. The weakest area opens at the frame that fails. The training log shows why, a gap goes to an owner, the medical monitor rehearses against the record, late filings get drafted notes, and Dana signs a storyboard with the meaning of her signature.

1

The war room, 19 days out

The war room is the solution's home screen. Dana starts from a countdown to the opening meeting: readiness 84 of 100, up 8 this week; 4 of 12 storyboards inspection-ready; 274 of 291 sentences backed; 14 open gaps.

The radar scores nine inspection areas. Protocol compliance at 72 and data handling at 76 sit below 80 and are marked in red. On the storyboard wall, each square is one frame of one story, colored by whether the record supports it. One card names the question an inspector is likely to ask next, the person who will answer it, and the fact that she has not rehearsed.

2

One click on the radar opens the frame that fails

Clicking protocol compliance opens SB-01, "Week 16 visits out of window at Site 114", at frame 6, What we did. The first sentence is cited: the site team retrained on March 20. The next is underlined in red: the new coordinator was trained before she booked her first visit.

Overnight the Claim Checker, the agent that tests each sentence, re-read all 31 claims in this storyboard: 30 supported, 1 not in the record. The red box gives the reason and the gap, G-01, owned by Omar Haddad and due October 14. The panel on the right holds the CAPA record itself, with the passages highlighted.

3

The training log shows why

The citation opens the Site 114 staff training log, read in place from the eTMF with nothing copied. K. Brooks has protocol training on March 3 and data system training on March 4. Her first visit-schedule training is the March 20 retraining, and she booked the Week 16 visit for participant 114-0019 on March 9.

The highlighted line says it plainly: no visit-schedule training is recorded for her before March 20. Omar can look for an earlier record, or the sentence gets reworded.

4

The answer, cited line by line, with coaching

Back in the war room, Dana opens the likely question: "Three Week 16 visits at Site 114 were out of window. How do you know your corrective action worked?" The answer draws only on the record, with a number on each sentence that opens its passage.

Underneath is coaching for Dr. Maya Chen, the medical monitor: lead with 9 of 9, do not say the coordinator was trained before March 9, and say "on-site visits" in place of "every visit". The coaching can only point toward what the record supports.

5

Dr. Chen rehearses, and the record checks what she says

She answers as she would in the front room. The Inspector Simulator, the agent that plays the inspector, holds her four sentences against the record. Two are backed. "The new coordinator was trained on the visit schedule before she booked any visit" is not in the record. "Our monitors check visit windows at every visit" is partly backed, since the plan covers on-site visits.

On the right is the back room, with its request log from the October 5 mock: four requests at 0:22, 0:38, 0:44 and 1:05, a median of 0:41. Each document is released by Omar.

6

Fix the record, or change the story

The gaps list shows 14 open across 8 storyboards, 4 of them high severity and 5 due by October 14, each with what is needed, the frame it blocks and an owner. Above them, the late-filing watch counts 37 documents filed October 5 to 7, 29 dated before July, 11 with a note-to-file.

One click drafts the 26 missing notes, each from the document's filing history with a factual reason. They then wait for Omar's signature, and nothing is filed until he signs.

7

Dana signs, and says what her signature means

SB-03, on an emergency unblinding at Site 209, has 21 sentences supported, 1 partly and none outside the record. Dana chooses the meaning, "Approved as factual and complete" or "Approved with the noted gaps", and signs with her password.

The partly supported sentence stays an open gap after she signs. In this setup, approving with a statement not in the record is switched off, so the dialog would send her back to reword the frame or close the gap first.

Week plan, data-change frame, back-room log and dashboard: the other inspection readiness screens

In shortBeyond the walk above: a frame with a timeline drawn from the record, the four-day session plan, the storyboards by stage, the data-change frame, the back-room log, the slide an inspector could receive, the dashboard, the approvals and rules, the sources read in place, and the same study seen for the MHRA.

These are design goals, to be measured in the live solution on your own study. They come from the fictional data on these screens. A readiness score of 84 is not a statement about how any inspection would go.

Six agents draft and check the storyboards; the QA lead signs and a person releases every document

In shortSix specialist agents index the record, propose topics, draft, check, rehearse and fetch. They never move or file a document and never release one. The maker and the checker stay apart: a QA lead signs each storyboard, the medical monitor reviews medical ones, and the TMF manager signs notes and releases documents.

Who signs an inspection storyboard, and who releases a document

Dana, Maya, Omar, Lena and Priya: the people in the VLT-302 screens

From the eTMF and quality system to signed storyboards with their gaps

What inspection preparation and a poor inspection cost your clinical QA team

In shortTwo costs sit apart. Preparation is certain and you can time it. A bad outcome is rare and heavy, and you cannot predict it. Published figures frame the second one. Your own storyboard hours, document search times and open gap counts size the first, and only you can measure them.

Timing one storyboard and ten document requests to find your own baseline

The repeat finding, the lost trust, and other inspection costs a calculator misses

What faster storyboards and document retrieval return when you count them against the tool's cost

In shortPut in your own numbers. Every default marked illustrative is an assumption to replace. The calculator values preparation days and back-room hours alone, and shows delay on a separate line because no one can promise that readiness work moved a date.

Storyboard days, back-room minutes and findings: where inspection automation can pay

Work out your inspection readiness return from your own counts

A sponsor with several programs under review, 12 storyboards each and two readiness exercises a year clears its cost at these defaults. A small sponsor with one study, inspected once in several years, would not pay this back on hours alone and should say so before anyone builds a business case. That sponsor may still choose to buy the protection, knowing that is the reason.

How ready your clinical QA team is to automate: storyboards, the record, retrieval, owners and sign-off

In shortPick the answer closest to your team today. Eight questions score 0 to 2, and the last is 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.

A low score means writing your storyboard rule before anything else

What to ask inspection readiness software about citations, gaps, coaching, your eTMF and signatures

In shortAsk every vendor the same questions, us included. The answers show what it does with a sentence that has no source, whether coaching stays inside the record, what it writes to your systems, and how you would measure it on your own closed study.

We found no published independent test of any inspection readiness tool as of November 2026. A planted-error test on your own closed study is the fairest comparison, and the same test measures your current process.

Rebuild one closed storyboard with planted errors, then run a mock with the back room on the clock

In shortStart with a storyboard you already signed. Plant errors your team has not seen, and see what gets caught. Then run a one-day mock with the back room on the clock. None of these steps commits you to a vendor, and each one tells you something about your record.

1. Time one storyboard and ten document requests, and export your filing lag for one study.
2. Write down your storyboard rule: the frames, the sources it may cite, and who may see it.
3. Take three closed topics, copy them and plant errors: a wrong date, a count rounded up, a sentence with no source, a superseded document cited.
4. Run the tool with read access only. Have a QA reviewer who has not seen the plants read the same text, then run a one-day mock with the back room timed.
5. Decide with four numbers: planted errors caught, false alarms per storyboard, citations that open the right passage, and the back-room median.

Planted errors, false alarms, citations and retrieval: how to score an inspection readiness pilot

GCP inspection readiness questions: BIMO notice, Form 483, storyboards, notes-to-file and mock inspections

In shortShort answers to what clinical QA teams search for, each one complete on its own. Rules and statuses are as of November 2026.