SearchManufacturing & Quality

When they ask

FDA inspection readiness, measured in minutes to the document

Readiness is usually bought as binders and rehearsals. What an investigator actually experiences is retrieval time — the gap between a request and the right page, at the right version, on the table. This guide puts numbers on both, from FDA's own FY2025 citation data, and shows what Batch Record Search changes about the back room.

Updated · 14 min read

See one case, screen by screen →
Watch with transcript →

FDA inspection readiness is usually prepared as a set of documents and a rehearsal, but the thing an investigator actually experiences is retrieval time: how long it takes the back room to put the right page, at the right version, in front of the person who asked. That gap is hard because a single request rarely maps to a single document — proving one process step means producing the executed lot record, the criticality analysis that classifies the parameter, the registered range in the dossier and the study that validated it, each owned by a different function and each with a version history. Botminds' Batch Record Search answers that kind of request from the records a site already keeps, citing the document, page and step for every claim, with a named reviewer checking at the page before anything is used.

The site examples on this page come from a fictional sample network inside the solution: three products made at Plants 1, 2, 3 and 5, with 2,598 batch records and 3,232 effective documents in English, Dutch and French. The inspection figures are FDA's own, from its published citation data and its Office of Pharmaceutical Quality reports, linked where they appear.

What FDA's own FY2025 drug citation data holds

713

Form 483s issued to the drugs programme in FY2025 from FDA’s inspection system

2,837

drug citations in that data, across 316 distinct citation rows

236

of them under 21 CFR 211.192, investigations of discrepancies and failures

972

FDA-conducted drug quality assurance inspections in FY2024, up 27% on FY2023

The first three are counts we computed from FDA's FY2025 Drugs worksheet, covering inspections ending between 1 October 2024 and 30 September 2025; they are not an FDA-published ranking. FDA notes the spreadsheets are not a complete list of observations, because some Form 483s are prepared by hand and never enter the system. The inspection count is FDA's own, from the FY2024 report on the state of pharmaceutical quality.Source: FDA, Inspection Observations, FY2025 dataset; FDA CDER, FY2024 State of Pharmaceutical Quality

What inspection readiness is, measured from the back room

Many readiness programmes are measured by what has been prepared: the site master file is current, the presenters are trained, the front room and back room are staffed, the tour route is agreed. Those are necessary and they are not the measure. The measure is what happens in the ninety seconds after an investigator asks for something — whether the back room knows which document answers, which version was in force on the date in question, who owns it, and how long it takes to put it on the table in a form the investigator can read.

That is worth stating as a number because it behaves like one. A site that cannot produce a lot record for a 2024 campaign inside an hour will not discover this during an inspection for the first time; it will discover it every time anyone asks, and the inspection is simply the occasion when the cost becomes visible. The useful readiness exercise is therefore not another mock inspection. It is a stopwatch on ten real requests, run on an ordinary Tuesday.

"All records required under this part, or copies of such records, shall be readily available for authorized inspection during the retention period at the establishment where the activities described in such records occurred."

21 CFR 211.180(c), on where records must be and how fast

What do FDA investigators cite most on drug inspections?

FDA publishes a spreadsheet of the areas of regulation cited on the Form 483s its inspection system generates, by fiscal year and by programme area. In FY2025, the drugs worksheet carries 2,837 citations across 316 rows. The six single rows cited most often are below, and they describe a programme of inspection that is mostly about written procedures, investigations and laboratory control.

The six most-cited single citations on US drug inspections, FY2025

211.22(d) · quality unit procedures not in writing or not followed
243
211.192 · investigations of discrepancies and failures
164
211.100(a) · absence of written production and process controls
162
211.160(b) · laboratory controls not scientifically sound
121
211.67(a) · equipment cleaning, sanitising and maintenance
95
211.68(b) · computer control of master formula records
87
Counted from FDA's FY2025 inspection observations data, Drugs worksheet, for inspections ending between 1 October 2024 and 30 September 2025. Each bar is one citation row; several rows can sit under the same regulation, so a section's total is higher than its largest row. FDA's own note applies: the data covers the Form 483s its system generates, not those prepared by hand.Source: FDA, Inspection Observations, FY2025 Drugs worksheet

Read as sections rather than single rows, the picture sharpens further. All the rows under 21 CFR 211.192, the production record review and discrepancy investigation requirement, add up to 236 citations — the second-largest section in the drugs data after the quality unit's own procedures. Under it sit rows named "written record of investigation incomplete" (32) and "no written record of investigation" (22). Those are not findings about a process going wrong. They are findings about the record of it.

The FY2025 numbers, and what they say about records

Three section totals in the same data are small enough to be worth looking at closely, because they are the ones that sit squarely on batch documentation. Citations under 21 CFR 211.188, the batch production and control record requirement, total 69. Under 211.186, master production and control records, 16. Under 211.180, general record requirements including retention and annual review, 64 — and six of those are the specific row "records not made readily available to FDA".

Six citations in a year is a small number, and it should be read carefully: the row counts records that were not made available, not records that arrived late. What the data cannot show, because no published dataset measures it, is how many requests were satisfied slowly: the hour in the back room, the call to a retired colleague, the superseded version produced first and corrected later. Those cost nothing in citations, and they still shape how the rest of the inspection goes.

The 236 citations under 211.192 are the more useful signal for a documentation programme. An investigation is only as good as the records it rests on, and an investigation that cannot cite the executed page, the parameter's criticality and the registered range is an investigation an investigator will push on. What makes a process parameter critical covers the classification side of that argument in detail.

How quickly must a record be produced?

There is no stated number of minutes. US drug GMP requires that records, or copies of them, be "readily available for authorized inspection during the retention period at the establishment where the activities described in such records occurred", and that they be available for photocopying or other reproduction as part of the inspection. The same paragraph adds the clause most relevant to a modern site: "Records that can be immediately retrieved from another location by computer or other electronic means shall be considered as meeting the requirements of this paragraph."

"Readily available" and "immediately retrieved" are the standard, and they are judged in the room. An investigator forms a view of a quality system partly from how a site behaves under a document request, and a back room that takes two hours over a lot record has told them something about the system before anyone reads the record.

Electronic records are squarely inside the request, too. FDA's data integrity questions and answers put it plainly: "All records required under CGMP are subject to FDA inspection. This applies to records generated and maintained on computerized systems, including electronic communications that support CGMP activities." The guidance gives an email authorising batch release as an example of a CGMP record FDA may review, and points back at 211.180(c) for the obligation to allow inspection, review and copying, including copying of electronic data.

The four documents behind one typical request

Take a request an investigator makes routinely: show me the hold before filtration on this lot, and show me it was inside what you registered. It sounds like one document. It is four, owned by four functions, and the two in the middle each have a superseded version sitting beside the effective one.

One request, unpacked: four documents, four owners, four version questions

What the investigator asksThe document that answersWho owns itWhich version

The executed record for the lot

The documentExecuted batch record, 318 pages, Dutch, Plant 3

OwnerPlant 3 document owner

VersionThe lot record as executed on 2 September

Is that step critical?

The documentProcess parameter criticality analysis CA-VLT-DSP-004

OwnerMSAT lead

Versionv3.0, effective 30 September — v2.0 retired

What did you register?

The documentCTD Module 3.2.S.2.4, the registered ranges

OwnerRegulatory CMC

VersionThe version in force for this market

How long is the hold validated for?

The documentHold time study HTS-VLT-012

OwnerProcess validation

VersionThe study the filing rests on

The fictional sample network, on one lot of a biologic drug substance. The two middle rows are where the version question bites: both documents have a superseded version sitting in the same system, and either will answer the question wrongly but plausibly.

In the sample network the answer to that request is that lot VLT-L3-26-0388 held its neutralized pool for 5 h 17 min before depth filtration, from 09:45 to 15:02 on 2 September at 21.8 °C, written at step 7.5.1 on page 58 of a 318-page Dutch record, with the operator's and checker's initials beside the entry. The batch record allows at most eight hours at 15–25 °C; the criticality analysis classifies the parameter as critical with a severity of 8 out of 10; the registered range in the dossier is the same eight hours; and the hold time study validated twelve. Four documents, one answer, and every number in it attributable to a page.

Version control under pressure: citing a superseded document

The quiet failure mode of a well-stocked back room is producing the wrong version confidently. In the sample network's records list, the Plant 3 master batch record exists as v5.2, effective 14 August 2026, and as v5.1, superseded — 412 and 409 pages respectively, both in Dutch, both findable, differing in three ultrafiltration steps under one change control. The criticality analysis exists as v3.0, effective 30 September, and v2.0, superseded, with two parameters reclassified between them.

An answer built on v2.0 of that analysis may classify a parameter as non-critical when the effective version calls it critical. If that answer reaches a deviation investigation, the investigation is wrong in a way that is both material and easy for an investigator to find, because the retired version is sitting in the same system. This is why the solution cites only effective versions unless a question names a past one, and why its intake agent retires superseded versions once the document's owner confirms the effective date rather than on upload.

A full answer in Batch Record Search, with its sources and their versions

A full answer screen headed AEX load conductivity at Plant 3 vs filing, flagged Differs from filing, asked by Lena Ortiz, 4.1 s. The answer reads: No. Plant 3 sets AEX load conductivity at 5.0 to 8.5 mS/cm at step 8.2.3, while the filed range is 5.0 to 8.0 mS/cm; the upper limit at Plant 3 is 0.5 mS/cm above the filing, and Plant 1 uses 5.0 to 8.0; raise this with Regulatory CMC before the next Plant 3 campaign. Below, the source step in its own language shows the Dutch original beside an English translation with the numbers unchanged. A plant-versus-plant table shows load conductivity 5.0 to 8.5 at Plant 3 page 188 step 8.2.3 against 5.0 to 8.0 at Plant 1 page 176 step 8.1.4, marked Differs, and load pH 7.2 to 7.6 at both, marked Same. A Sources panel on the right lists three documents with plant, version, language, page and step, each with the cited sentence and a link to open at that step.
Fictional sample network. Each of the three sources carries the version it was read at, so the answer can be re-checked at the page rather than taken on trust.

What the inspection volume looks like now

Inspection volume has climbed steeply since the pandemic. FDA's Office of Pharmaceutical Quality reports that in FY2024 there were 972 FDA-conducted drug quality assurance inspections, an increase of 27% from the 766 accomplished in FY2023, which was itself up from 522 in FY2022. At the end of FY2024, CDER's site catalog held 4,619 manufacturing sites worldwide, 41% of them in the United States.

The geography matters for multi-site networks. The same report records that "In FY2024, more than 62% of drug quality assurance inspections were at foreign sites —an all-time high", with 34% of catalogued sites in India and 28% in China inspected that year, against 24% of US sites. For a network whose records are written in three languages at four plants, that means the inspection is as likely to land on the site whose records nobody at headquarters can read as on the one next door.

Running the back room as a retrieval desk

The back room's job during an inspection is identical to the job an MSAT or quality desk does every other week of the year: find the document, confirm the version, read the step, state what it says. Treating it as a different activity — staffed by different people, using a different filing arrangement, rehearsed rather than practised — is what makes it slow.

The back room as many sites staff it, against the back room as a retrieval desk

Staffed for the week of the inspectionRun as the desk that answers all year
Who is in it

Staffed for the week of the inspection: Senior people pulled off their own work for the duration, plus a runner.

Run as the desk that answers all year: The people who answer document questions every week, with the same tools they use every week.

How a request is resolved

Staffed for the week of the inspection: By knowing who to call. The knowledge lives in a handful of long-serving heads.

Run as the desk that answers all year: By searching the document system the records already live in, and reading the cited page.

Version control

Staffed for the week of the inspection: Checked by the person who fetches, under time pressure, at the moment of maximum cost.

Run as the desk that answers all year: The effective version is what comes back; a superseded one appears only when the request names it.

What is handed over

Staffed for the week of the inspection: A document, with the relevant part found by the investigator or pointed to verbally.

Run as the desk that answers all year: The page, with the step identified, and the other three documents the request implies already located.

What is left behind

Staffed for the week of the inspection: A list of commitments reconstructed from notes afterwards.

Run as the desk that answers all year: A record of every request, what was produced, which version, and who checked it.

How it is measured

Staffed for the week of the inspection: By the number of observations at the close-out meeting.

Run as the desk that answers all year: By time to the right page, on ordinary requests, all year.

The right-hand column is how the work is arranged when the back room uses the same answering desk the site runs between inspections. The left-hand column is a common arrangement, not a regulatory requirement.

One part of the right-hand column is not optional, and it is the part a tool must not take over. Under US drug GMP, "all drug product production and control records, including those for packaging and labeling, shall be reviewed and approved by the quality control unit to determine compliance with all established, approved written procedures before a batch is released or distributed". Retrieval can be automated. Review and approval cannot, and in the solution the signature that carries them is a named person's, with its meaning recorded and the cited versions kept beside it.

Signing an answer in Batch Record Search: what it means, and where it may be used

A Verify this answer dialog over the AEX load conductivity answer. It summarises the answer as 3 citations, a comparison table and 3 checks. A field headed Meaning of this signature reads Verified — the cited steps match the source documents. A field headed Use in reads Save as a verified answer for the team. A Signed by line reads Dr. Maya Chen, Quality reviewer, identity confirmed by single sign-on, Oct 6 2026. A note says the electronic signature, its meaning and the cited versions are kept in the audit trail. Buttons read Cancel and Sign and verify.
Fictional sample network. The signature records what it means — that the cited steps match the sources, or that the answer is fit for use in a deviation investigation or a change control — so a later reader knows what was attested to.

A readiness test you can run on your own records this month

The exercise below needs one morning, two people and no new software. Pick ten requests of the shapes investigators actually make, run them cold on real documents, and record the clock and the version. The time to beat is not a published figure; it is whatever the slowest of the ten took, which is the one an investigator will find.

Ten requests, timed cold, on your own documents

  1. The executed record for a named lot from two years ago

    A routine request. Time it from the request to the page open, not to the file located.

  2. One process step inside that record, by name rather than by step number

    Investigators ask for the hold before filtration, not for step 7.5.1. Someone has to make that translation.

  3. The version of the master record that was in force on the lot’s manufacturing date

    Not the current one. This is where a site can produce the wrong document confidently.

  4. The criticality classification of the parameter in that step

    And the version of the analysis it comes from, with its effective date.

  5. The registered range for the same parameter in the dossier

    Owned by regulatory affairs, usually in a different system from the batch records.

  6. The study the validated limit rests on

    Hold time, cleaning validation or process validation, as applicable to the step.

  7. The deviation record for a lot that had one, and the batch record page it refers to

    The pairing is the point: 236 FY2025 drug citations sit under the investigation requirement.

  8. The same step at a second site that makes the same product

    If the two differ, you want to know before the question is asked in the room.

  9. A record in a language the person answering does not read

    Count the time to a translation that keeps the numbers, units and step numbers exactly as written.

  10. Proof that the second-person check was performed on the entry

    Initials, times and the procedure that required them. An unsigned entry is a finding waiting to be made.

Run the ten, write the times down, and the readiness conversation changes shape: instead of arguing about how many rehearsals to hold, a site is looking at which of ten ordinary questions takes forty minutes and why. For many sites the answer is not the document system but the step between the request and the document — the knowing which one, at which version. Why a document system stores well and finds badly takes that apart, and what good documentation practice expects of an entry covers the handwriting the back room has to read. To see a request answered end to end on sample records, start from Batch Record Search.

The estate a back room has to reach into, version by version

A Records screen headed Read where they live in your document system — nothing is copied or migrated, 3,232 effective documents across four plants. Filters read All, Master records, Executed lots, Criticality analyses, Filings, Other, with a language filter. Rows list: Veltrimab drug substance batch record purification, master batch record, Plant 3, Dutch, v5.2 of Aug 14 2026, Effective, cited in 5 answers, 412 pages, owner Lena Ortiz; the same record v5.1 of Feb 3 2026, Superseded, 409 pages, 1 answer; the Plant 1 record v7.0 of Jun 2 2026, English, 386 pages, Effective; the Plant 5 record v3.1 of Jul 9 2026, French, 398 pages, Effective; a Corventa prefilled syringe record at Plant 2, English, v5.0, 164 pages; a Talmirex tablets record at Plant 5, French, v2.4, 96 pages; and a Veltrimab downstream process parameter criticality analysis, all plants, English, v3.0 of Sep 30 2026, Effective, cited in 4 answers, with v2.0 of Nov 18 2025 marked Superseded.
Fictional sample network. Each row carries the version, its effective date and its status, and the count of answers that cite it — so a back room can see at a glance which document an inspection is most likely to reach.

Questions people ask

What is a Form 483?

A list of observations an FDA investigator leaves at the end of an inspection. In FDA's words, observations "are listed on an FDA Form 483 when, in an investigator's judgment, the observed conditions or practices indicate that an FDA-regulated product may be in violation of FDA's requirements". It is not a final agency determination; the site responds, and what follows depends on the response and the severity. FDA's system issued 713 of them to the drugs programme in FY2025.

How quickly must you produce a record in an inspection?

No regulation states a number of minutes. The requirement is that records be readily available for authorised inspection at the establishment where the activity happened, and 21 CFR 211.180(c) treats records that "can be immediately retrieved from another location by computer or other electronic means" as meeting it. In practice the standard is judged in the room, which is why timing ten ordinary requests tells a site more than another rehearsal.

Can FDA look at electronic records?

Yes, and at the systems holding them. FDA's data integrity guidance states that all records required under CGMP are subject to inspection, including those generated and maintained on computerised systems and electronic communications that support CGMP activities — it gives an email authorising batch release as an example. The same guidance points back at 211.180(c) for the obligation to allow inspection, review and copying of electronic data.

Which part 211 sections are cited most on drug inspections?

In FDA's FY2025 drugs data, the single most-cited citation is 21 CFR 211.22(d), quality unit procedures not in writing or not fully followed, at 243. Next come 211.192 on investigations of discrepancies and failures at 164, 211.100(a) on written production and process controls at 162, and 211.160(b) on laboratory controls at 121. Counted as whole sections rather than single rows, 211.192 reaches 236.

How many FDA drug inspections are there a year?

FDA's Office of Pharmaceutical Quality reported 972 FDA-conducted drug quality assurance inspections in FY2024, up 27% from 766 in FY2023 and 522 in FY2022, plus 198 inspections classified under mutual recognition agreements. More than 62% of the FY2024 inspections were at sites outside the United States, the highest share recorded.

What happens if you cannot find a document during an inspection?

If a required record genuinely cannot be produced, it is citable: FDA's FY2025 drugs data carries six citations under the row "records not made readily available to FDA". The more common outcome is slower and less visible — the record arrives late, or the wrong version arrives first. Neither is a citation on its own, and both change how the rest of the inspection goes.

Batch Record Search.
See it on your own files.

An agentic solution from Botminds. Ask about any lot or step; the answer comes back beside the scanned page, the step boxed and marked if it is critical. Book a demo, or explore how it works.