Picture the night shift at a sterile fill line. At 02:10 the operator adds 40 mL of buffer to bring a solution into range, a step the master record does not list. The electronic batch record (EBR) has no field for it, so the operator writes it on a sticky note and moves on. At 08:00 the batch record looks perfect. Six months later an investigator compares the record with the logbook, finds the gap, and writes it up: a deviation from the written procedure, neither recorded nor justified.

Electronic batch record requirements come from two sections of FDA's drug CGMP rules. 21 CFR 211.100(b) says production steps must be documented at the time of performance, and any deviation must be recorded and justified. 21 CFR 211.188 lists what each batch record must contain, from equipment IDs to the person who checked each significant step. If you build MES, EBR or electronic logbook software for drug manufacturers, those two sections are your spec.

Electronic batch record requirements under 21 CFR 211.188(b) mapped to what an EBR system must capture

What must an electronic batch record contain?

Section 211.188(b) requires documentation that each significant step was accomplished. Read it as a schema:

211.188(b) Requirement What your EBR stores
(1) Dates Server timestamps per step
(2) Major equipment and lines used Equipment ID, scanned at the step
(3) Each batch of component or in-process material Lot IDs, scanned, matched to the bill of materials
(4) Weights and measures Values from the balance or meter, not typed
(5) In-process and lab control results Results with limits and pass or fail
(7) Actual yield and percent of theoretical Calculated per phase, with limits
(10) Any sampling performed Sample IDs, quantities, who sampled
(11) Who performed and who checked each significant step Two distinct user IDs per critical step
(12) Any investigation under 211.192 Link to the deviation or investigation record

Section 211.188(a) adds one more: an accurate, checked, dated and signed reproduction of the master production record. In software terms, the batch record must point at a locked, approved version of the recipe.

How often do FDA 483s cite batch records and production procedures?

Often. We looked at 212 Form 483s issued to drug, biologic and animal-drug facilities in 2025 and 2026 in the Atlas Form 483 directory, the same sample as our post on 21 CFR Part 11 audit trail requirements.

  • 27.8% cite 21 CFR 211.100, written production procedures and deviations.
  • 9.0% cite 211.110, in-process controls.
  • 7.1% cite 211.188, batch production and control records.
  • 33.5% cite 211.100, 211.186 or 211.188, one 483 in three.

Those 483s run long. They average 6.5 observations, against 4.3 across all 212.

Share of 2025 and 2026 drug Form 483s citing 21 CFR 211.100, 211.110 and 211.188, with average observations per 483

Which real 483s cited batch records and deviations?

Each link opens the public observation list.

Batch records incomplete (211.188):

Deviations and procedures (211.100):

A two-observation 483 can carry this finding. Safecor and Sciegen each received two.

How do you build deviation capture into an EBR?

Make the off-script path as easy as the scripted one. The sticky note in the story exists because the system offered no field for what happened. Every step needs an "unplanned event" action that records what happened, when and who, blocks the step from closing until someone justifies it, and routes it to QA.

from dataclasses import dataclass, field from datetime import datetime, timezone def now() -> datetime: return datetime.now(timezone.utc) # server clock, never the client's  @dataclass class Deviation: step_id: str description: str raised_by: str raised_at: datetime = field(default_factory=now) justification: str = "" qa_approved_by: str | None = None @dataclass class Step: step_id: str critical: bool equipment_id: str | None = None performed_by: str | None = None checked_by: str | None = None completed_at: datetime | None = None deviations: list = field(default_factory=list) def complete(self, performed_by: str, checked_by: str | None = None): if self.critical and (not checked_by or checked_by == performed_by): raise ValueError("211.188(b)(11): critical step needs a second, different person") if any(not d.justification or not d.qa_approved_by for d in self.deviations): raise ValueError("211.100(b): deviation not yet justified and approved") self.performed_by, self.checked_by, self.completed_at = performed_by, checked_by, now() def yield_pct(actual_kg: float, theoretical_kg: float) -> float: return round(100 * actual_kg / theoretical_kg, 1) # 211.188(b)(7), checked against 211.103 limits  s = Step("FILL-030", critical=True) s.deviations.append(Deviation("FILL-030", "Added 40 mL buffer to bring pH into range", "op.lee")) try: s.complete("op.lee", "sup.ortiz") except ValueError as e: assert "211.100(b)" in str(e) s.deviations[0].justification, s.deviations[0].qa_approved_by = "pH adjustment per SOP-118 rev 4", "qa.hall" s.complete("op.lee", "sup.ortiz") assert yield_pct(96.4, 100.0) == 96.4

Three rules sit under that code:

  • Capture at the time of performance. 211.100(b) says "documented at the time of performance". Disable backdating, stamp steps with server time, and show late entries as late.
  • Read values from instruments. Weights typed by hand invite transcription errors. Pull them from the balance and keep the raw reading.
  • Lock the recipe. The batch record should reference one approved master record version. A recipe edit mid-batch is a change control event, not a field update.

What else does an FDA-ready EBR need?

Batch records are electronic records under 21 CFR Part 11, so the audit trail and e-signature rules apply. Our Part 11 guide covers that design, and the audit trail tests used in computerized system validation list what a reviewer will try. For validation effort, computer software assurance shows how to scale testing to risk. On the quality side, whether pharma companies are ready for FDA eLog and eBatch audits and the documentation failures FDA flagged in 2024 and 2025 show what reviewers find.

EBR checklist for engineering teams

  1. Every 211.188(b) element has a field, and the batch cannot close with one empty.
  2. Critical steps require two different users: one performs, one checks.
  3. Every step offers a deviation action that blocks closure until QA approves the justification.
  4. Steps are stamped with server time; late entries are flagged, never hidden.
  5. Weights, measures and equipment IDs come from devices or scans, not keyboards.
  6. Yield is calculated per phase and compared against the limits in the master record.
  7. The batch record points at one locked, approved master record version.
  8. Audit trail and e-signatures meet Part 11.

FAQ

What are the electronic batch record requirements under FDA rules?
21 CFR 211.188 requires a batch record for each batch that reproduces the approved master record and documents each significant step: dates, equipment, component lots, weights and measures, in-process results, yields, labeling, sampling, the people who performed and checked each step, and any investigations. 21 CFR Part 11 adds audit trail and electronic signature controls when the record is electronic.

What does 21 CFR 211.100(b) require?
Written production and process control procedures must be followed and documented at the time of performance. Any deviation from them must be recorded and justified. In an EBR, that means real-time step capture and a deviation workflow with justification and approval.

How often do FDA 483s cite batch records?
In 212 drug, biologic and animal-drug Form 483s from 2025 and 2026, 7.1% cited 21 CFR 211.188 and 27.8% cited 211.100. One in three cited 211.100, 211.186 or 211.188.

Does an electronic batch record need two signatures?
For each significant step, 211.188(b)(11) requires the identity of the person who performed it and the person who directly supervised or checked it. If automated equipment performs the step, the record must identify the person who checked it.

Is an EBR system subject to validation?
Yes. 21 CFR 211.68 requires that computer systems used in production are checked and controlled, and Part 11.10(a) requires validation of systems that hold electronic records. FDA's computer software assurance approach lets you scale testing to the risk of each function.

I work on Atlas Compliance, which tracks FDA inspections and Form 483 observations for pharma and medical device quality teams. The FDA inspection directory is free to browse.