In‑Process Control Checks (IPC) – Controlling Quality While You’re Still Making the Product

This topic is part of the SG Systems Global regulatory & operations glossary.

Updated November 2025 • BMR/eBR | MES | SPC | CPV | PAT | Weighing & Dispensing | Yield Variance | Batch Variance Investigation | Deviation/NCR | CAPA | Data Integrity •
QA, QA Ops, Manufacturing, Tech Ops, CI, Digital

In‑process control checks (IPC) are the measurements, inspections and tests performed during manufacture – not only at raw‑material receipt or finished‑product release – to verify that the process is operating within its defined parameters and will deliver compliant product. IPCs cover everything from simple line checks (weights, labels, temperatures) to sophisticated inline PAT measurements and real‑time quality predictions.

In a modern plant, IPCs are where your control strategy either becomes real or gets exposed as wishful thinking. They are the practical expression of “state of control”: did the dough actually hit the target temperature? Did the bake profile stay within window? Did viscosity, potency or moisture behave as the design space says it should? If you only find out at final QC that something went wrong, your IPC regime is not doing its job.

“End‑product testing tells you what you made. In‑process control checks tell you whether you were in control while you were making it – which is what regulators actually care about.”

1) What We Mean by In‑Process Control Checks

In‑process control checks sit between incoming inspection and final release testing. They are performed while material is being transformed, not just when it arrives or leaves the plant. Typical flavours:

• Parameter checks: line speed, temperature, pressure, mixing time, proof time, hold time, vacuum, humidity, etc.
• In‑process quality checks: weight, fill level, pH, viscosity, density, moisture, texture, colour, dimensions, appearance.
• Functional checks: tablet hardness, seal integrity, torque, adhesion, rise and spread in dough, crust colour, crumb structure.
• Compliance checks: label correctness, code correctness, allergen status, cleaning verification, segregation and line clearance.

IPCs may be manual (operator readings with a balance or pH meter), semi‑automated (checkweighers, vision systems) or fully automated PAT instruments feeding closed‑loop control. They can be destructive (taking samples) or non‑destructive (inline sensing). The common thread is:

• They are planned, documented and part of the control strategy.
• They have acceptance criteria and defined actions for failure.
• They are recorded as part of the batch record and inform batch disposition.

If you measure something “because we’ve always done it” with no clear purpose, criteria or action, that’s not IPC – that’s busywork masquerading as control.

2) Why IPCs Matter – Beyond End‑Product Testing

It is tempting to trust final QC: if the product passes release tests, the process must be fine, right? Wrong – at least from a regulatory and risk point of view. IPCs matter because:

• They give earlier warning:
  • Dough temperature, viscosity, density, or intermediate potency can drift long before final results fail. IPCs catch issues when you can still adjust or quarantine cheaply.
• They prove “state of control”:
  • Regulators expect you to demonstrate that the process was controlled, not just that a few final samples were within spec. IPC data is the evidence.
• They reduce scrap, rework and yield variance:
  • Adjusting earlier in the process beats discovering a train wreck at the end. Every time. See Yield Variance and Batch Variance Investigation.
• They make investigations faster:
  • When something does go wrong, good IPC data shows where and when the process left normal behaviour instead of sending you on a fishing expedition.
• They unlock real process understanding:
  • Continuous IPC data feeding CPV tells you which parameters actually matter, not just which ones looked important on a whiteboard during validation.

A plant that leans only on final testing is basically gambling: as long as the dice keep rolling your way, it feels fine. IPCs change the game from gambling to control.

3) Regulatory & Standard Expectations for IPC

Across pharma, food, and other regulated sectors, guidance converges on the same principles:

• Key process parameters and in‑process attributes should be identified and controlled based on science and risk (QbD, control strategy).
• Monitoring and control during production are essential; not everything can be “tested into compliance” at the end.
• IPC results that may impact quality or safety must be recorded and reviewed as part of the batch record.
• Non‑conforming IPC results must trigger evaluation, investigation and, where appropriate, batch impact decisions and CAPA.

In practice, regulators look for:

• A documented IPC plan per product or process, aligned with validation and risk assessments.
• Clear links between IPCs and CPPs/CQAs defined in your control strategy and process validation/PPQ.
• Evidence that IPCs are actually performed, within defined frequency and tolerances, and properly documented in the BMR/eBR.
• Robust handling of IPC failures – not just “we adjusted and carried on” with no trace.

“We rely on good operators; they’ll spot if something looks wrong” is not a control strategy. It’s an admission that you haven’t turned process knowledge into enforceable checks.

4) What IPCs Typically Cover

Good IPC design doesn’t measure everything; it measures the right things. Typical IPC categories:

• Critical process parameters (CPPs):
  • Mixing time and speed, jacket temperature, proofing temperature and humidity, oven zones, drying temperature, critical agitation or aeration rates, line speeds.
• In‑process quality attributes:
  • pH, viscosity, density, particle size, moisture content, intermediate assay, dough rheology, gas cell structure, crust colour, texture.
• Weights and counts:
  • Fill weights, piece counts, tablet weights, loaf weights, dose volumes – often controlled by checkweighers and sampling plans.
• Visual and sensory checks:
  • Appearance, colour, defects, foreign matter, odour, taste panels (for example, Finished Product Sensory Evaluation (Baking)).
• Compliance & identity checks:
  • Label and code verification, component ID checks, allergen status, cleaning verification, line clearance, correct tooling or pans.
• Yield & loss indicators:
  • Scrap and giveaway, moisture loss vs plan (see Moisture Loss and Bake Yield Testing), unusual trim, rework levels.

A helpful rule of thumb: if a parameter or attribute is critical enough to feature in your control strategy or risk assessments, it probably needs some form of IPC – either direct measurement or a justified surrogate. If it never shows up in any IPC, either it isn’t truly critical or your design has a hole.

5) Designing IPC Plans – Risk‑Based and Coherent

IPC design should be deliberate, not just a copy‑paste of whatever checks the last project used. Key design steps:

• Start from process understanding:
  • Use QbD work, DoE, lab and pilot data, tech transfer reports and QRM outputs to identify where the process is most sensitive.
• Map CPPs and CQAs to IPCs:
  • For each CPP/CQA, decide how you will monitor it in real time: direct measurement, surrogate, or justified reliance on upstream/downstream controls.
• Define acceptance criteria and actions:
  • For each IPC: target, limits, frequency, what to do on fail, who decides, and how results are recorded.
• Level of automation:
  • Decide which IPCs should be manual (low frequency, low risk), which should be semi‑automated, and which should be continuous/inline.
• Integrate with EWIs and eBR:
  • IPCs need to appear in work instructions, MES phases and eBR steps with prompts, limits and enforced recording.
• Limit to what you will act on:
  • If you have no intention or ability to act on a measurement, don’t add it “for information”. You are only creating noise and extra work.

A good IPC plan is not a long checklist. It is a focused set of checks that are clearly linked to risk and to specific actions when the results go wrong.

6) Sampling, Frequency & Statistical Considerations

IPCs are not just about what you measure, but how often and with what coverage. Design choices:

• Continuous vs periodic:
  • Critical, fast‑moving parameters (oven temperature, line speed) are typically monitored continuously, with alarms and interlocks.
  • Attributes like weight, dimensions or appearance may use periodic sampling (for example, every 15 or 30 minutes, X pieces per check).
• Sampling basis:
  • By time (every 30 minutes), by volume (every N kg), by units (every N packs), by batch stage or by event (start up, after changeover, after intervention).
• Statistical thinking:
  • For attributes, define sample size and criteria (for example, acceptance numbers, AQL logic, control charts), not “check a few and hope”.
• Dynamic frequencies:
  • Some IPCs can become less frequent once the process is proven stable, with increased sampling when CPV or SPC shows new risk emerging.
• Start‑up and changeover IPCs:
  • Heavier checking around start‑up, recipe changes, allergen changeovers and equipment maintenance is usually justified.

“Check one pack per hour” is not a control strategy; it’s ritual. IPC sampling plans should be defensible to a statistician, not just comfortable for an operator.

7) Execution – How IPCs Actually Happen on the Line

On paper, IPCs are bullet points in a control strategy. In reality, they are tasks, tools and behaviours on the shop floor. Execution design matters:

• Who performs IPCs?
  • Typically line operators or technicians for routine IPCs; QA or lab staff for specialised tests or confirmatory checks.
• Where is the method?
  • Methods must be accessible and clear: how to sample, how to operate the instrument, how to interpret results, what to do on fail.
• How is data captured?
  • Best: directly into MES/eBR, checkweighers, PAT systems or tablets – no re‑transcription from scraps of paper.
  • At worst: controlled paper forms transcribed quickly into the batch record with checks for transcription errors.
• What happens on fail?
  • Action trees should be simple and explicit: stop, segregate, adjust, re‑sample, call QA/Engineering, open deviation, etc.
• How is training maintained?
  • IPCs are only as good as the people doing them. Competence, not just attendance at training, has to be demonstrated.

If operators describe IPCs as “extra paperwork QA makes us do” rather than “how we keep the line in control”, you have a culture problem and a design problem, not just a training gap.

8) Digital IPC – eBR, MES, PAT & Automation

Digitalisation is what makes IPC scalable and reliably enforceable. Elements of a good digital IPC setup:

• MES/eBR‑driven checks:
  • IPCs embedded as steps in the electronic batch record, with prompts, limits, and enforced capture before the process can continue.
• Inline sensors and PAT:
  • Near‑infrared, spectroscopy, vision systems, checkweighers, temperature and humidity networks feeding live data instead of intermittent manual readings.
• SPC and alarms:
  • Automatic SPC charts and alarms when trends indicate loss of control – not just post‑hoc charting for reports.
• Exception‑based review of IPC data:
  • Use Exception‑Based Process Review so QA focuses on IPCs that matter: fails, trends, missing checks, not every normal reading.
• Integration into analytics and CPV:
  • Feed IPC data into a GxP data lake for cross‑batch analysis, CPV, yield optimisation and debottlenecking.

Slapping tablets on the line doesn’t make IPC “digital” if people still write results on paper and key them in later. The objective is straight‑through, validated data capture with minimal manual handling and maximum built‑in logic.

9) Handling IPC Failures – What Happens When Checks Fail?

IPC failures are not rare events; if they are, your limits are probably too loose. The question is what you do when they occur.

• Immediate containment:
  • Stop or slow the process, segregate suspect material, prevent further mixing of conforming and non‑conforming product.
• Confirm and diagnose:
  • Repeat the measurement (without abusing re‑testing), check the instrument, review recent trends and events.
• Decide on adjustment vs rejection:
  • If within defined rules, you may adjust set‑points and continue operation, with documented rationale and confirmation samples.
  • In other cases, you may need to reject or downgrade material, or move it into investigation status.
• Trigger deviations where appropriate:
  • Certain IPC failures – especially CPP breaches, repeated out‑of‑control signals or data‑integrity concerns – should automatically trigger a deviation/NCR.
• Assess batch impact:
  • Evaluate how much material was affected, whether later steps can recover quality, and whether existing data supports release or requires rejection.

In a healthy system, repeated IPC failures drive CAPAs, engineering actions or recipe changes. In a broken system, they simply lead to “adjust and carry on” with increasingly creative justifications in the batch record – until a regulator or customer calls you on it.

10) IPC Data, Trending & CPV

IPC data is not only for real‑time control; it’s the backbone of long‑term process understanding.

• CPV and annual reviews:
  • Continued Process Verification relies heavily on IPC trends: are CPPs stable, are CQAs drifting, are yields or scrap creeping upwards?
• Correlation analysis:
  • IPC datasets allow you to correlate parameters with quality outcomes, revealing which levers truly drive performance.
• Control‑limit refinement:
  • Over time, you may tighten or adjust IPC limits based on real performance, or relax needlessly tight limits that generate noise without improving quality.
• Batch variance investigations:
  • IPCs are often the first place you see the story of a poor batch. Stable IPCs with a bad final result point to lab issues or late‑stage problems; wild IPCs point to process control issues. See Batch Variance Investigation.
• CI and debottlenecking:
  • IPC data shows where conservative limits are throttling throughput, where changeovers create instability, and where equipment capability exceeds current use.

If IPC results vanish into archives once a batch is released, you’re throwing away one of the most valuable datasets you have for process improvement and risk reduction.

11) IPC in Bakeries and Food Plants

In bakery and broader food operations, IPCs are often the only realistic way to control quality at industrial speeds. Typical examples:

• Dough preparation:
  • Flour protein/ash checks vs supplier COA; water temperature; target dough temperature; dough absorption and consistency; preferment age; see Flour Protein and Ash Variability Control, Target Dough Temperature Control and Dough Absorption Control.
• Fermentation and proofing:
  • Proof time, temperature and humidity windows; dough volume increase; poke tests; Proofing Validation (Dough Development).
• Baking:
  • Oven profile (zone temperatures vs set‑points), bake time, moisture loss, internal crumb temperature at end of bake; see Bake Profile Verification and Moisture Loss & Bake Yield Testing.
• Crust & crumb quality:
  • Crust colour uniformity, crumb texture profile, cell structure; see Crust Color Uniformity Testing and Texture Profile Analysis (Bakery Crumb Quality).
• Weights, labelling and allergens:
  • Piece weights against legal minimum, label correctness, date codes, allergen changeover swabs; see Allergen Changeover Verification (Bakery).

Large bakeries that treat IPCs casually end up relying on destructive end‑of‑line checks that cover a tiny fraction of daily output. That might scrape by for commodity bread; it’s a lot less amusing when a retailer’s private‑label range is on the line and every mis‑labelled pack is a potential recall.

12) Roles & Responsibilities

Designing and running IPCs is cross‑functional by nature.

• Technical Services / MS&T:
  • Define IPCs based on process knowledge, QbD and validation; specify methods, limits and criticality.
• QA / QA Ops:
  • Ensure IPCs cover risk appropriately; approve IPC plans; define how IPC results feed into batch release and deviations; audit execution and data integrity.
• Operations / Production:
  • Execute IPCs, respond to results, maintain instruments, escalate failures and provide practical feedback on feasibility and timing.
• Laboratory / QC:
  • Support specialised in‑process analyses, maintain and validate IPC methods and instruments, trend data.
• Engineering & Automation:
  • Integrate sensors and devices, ensure robust data capture and alarming, maintain equipment capability.
• Digital / Data teams:
  • Integrate IPC data into MES/eBR and analytics platforms, enable SPC, CPV and exception‑based process review.

When IPC design is left solely to QA, you tend to get long, impractical checklists that operations quietly bypass. When it’s left solely to Operations, you tend to get the bare minimum needed to keep lines moving. The right answer sits in the tension between those extremes.

13) Common Failure Modes & Audit Findings

Weak IPC systems usually fail in predictable ways:

• Tokenistic checks:
  • IPCs added to satisfy a template, not because anyone believes they control risk. Results are recorded but never used.
• Unclear actions on failure:
  • Operators know how to take measurements but have no idea what to do when results fail – or they improvise, batch by batch.
• Inconsistent execution:
  • IPCs skipped when the line is busy, done late or in batches to catch up, often with suspiciously neat numbers.
• Data‑integrity issues:
  • Results entered long after the fact, with no time stamps, corrections with no reason codes, or obviously “back‑filled” numbers.
• No link to control strategy:
  • IPCs check trivial things while critical parameters identified in validation mysteriously have no in‑process monitoring.
• No trending or learning:
  • IPCs are treated as batch‑by‑batch hurdles, not as part of CPV or yield improvement.

Auditors spot these patterns quickly. When they see IPC forms immaculate but obviously written with the same pen at the end of the shift, or see CPPs from validation absent from any IPC plan, they infer that your “state of control” is more fiction than fact.

14) Implementing or Upgrading an IPC Programme

Strengthening IPC doesn’t require a big‑bang project. A pragmatic path:

• 1. Map what you actually do today:
  • List current IPCs by product and line; note who does them, how often, with what tools and what actions on fail.
• 2. Compare to control strategy:
  • Check whether identified CPPs and CQAs have corresponding IPCs; highlight gaps and pointless checks.
• 3. Fix the worst mismatches first:
  • Add IPCs where risk is high and coverage is poor; remove or simplify IPCs that generate work without controlling anything meaningful.
• 4. Strengthen methods and data capture:
  • Standardise instruments, procedures and data capture routes; move away from uncontrolled spreadsheets and scraps of paper.
• 5. Integrate with MES/eBR and SPC:
  • Embed IPCs into digital workflows; implement basic SPC where variability matters; set up simple exception‑based review for IPC failures and trends.
• 6. Use IPC data in CPV and CI:
  • Pull IPC data into annual reviews, CPV and CI projects; demonstrate how better IPCs directly reduce scrap and complaints.

If you can’t clearly articulate, for a high‑volume product, “these are our in‑process checks, here’s why we do them, here’s what we do when they fail, and here’s the trend over the last year”, your IPC programme still has work to do.

15) FAQ

Q1. What’s the difference between in‑process control checks (IPC) and final product testing?
Final product testing looks at a relatively small number of samples at the end of the process to confirm that the batch meets specifications. IPCs monitor the process and intermediate material while production is happening, often continuously or at high frequency. IPCs are about keeping the process in control and catching problems early; final testing is about verifying the result. Regulators expect both, but they put increasing weight on whether your process, as demonstrated by IPC data, was actually under control.

Q2. How do we decide which IPCs are “critical”?
Critical IPCs are those that monitor or directly control critical process parameters (CPPs) or critical quality attributes (CQAs) – i.e. failure could impact patient safety, product quality or regulatory commitments. Use QbD studies, risk assessments and validation data to identify which parameters significantly influence CQAs, and classify IPCs accordingly. Critical IPCs deserve tighter method control, more robust data capture, higher automation and stronger escalation rules than routine, low‑risk checks.

Q3. Can we replace manual IPC sampling entirely with PAT and automation?
In some cases you can move from manual grab samples to inline or online PAT measurements, but that is a method change that must be justified, validated and reflected in your control strategy. Automation can dramatically improve coverage and responsiveness, but it introduces its own risks (sensor drift, calibration, data integrity, model validity). A mature setup often blends both: PAT for continuous control and early warning, supported by periodic manual or lab checks to verify performance and provide backup if the PAT channel fails.

Q4. Does every failed IPC result require a deviation?
Not necessarily. Your SOPs should distinguish between IPC failures that always trigger a deviation (for example, CPP outside validated range, clear data‑integrity concern) and those that can be handled within a defined adjustment and re‑check procedure with documented rationale. However, repeated “minor” failures in the same area, or any case where batch impact is unclear, should escalate into formal deviation and potentially CAPA. If you find yourself explaining the same IPC failure pattern over and over again without root‑cause work, you’re gaming the system.

Q5. What’s a practical first step to improve IPC without buying new systems?
Start by cleaning up what you already have. For one important product or line, map existing IPCs, throw out checks that nobody uses to make decisions, tighten or clarify acceptance criteria and actions for the ones that matter, and make sure results are recorded consistently in the batch record. Then put very simple trending in place – even if it’s just a basic control chart – and use that in weekly reviews. Once you’ve proved that better‑designed IPCs reduce headaches and scrap, it’s much easier to justify investment in digital capture, PAT and exception‑based review.

Related Reading
• Process & Control: ISA‑88 | MES | eBR | PAT | SPC
• Quality, Risk & Verification: CPV | Batch Variance Investigation | Deviation/NCR | CAPA | QRM | Data Integrity | Product Quality Review (PQR/APR)
• Bakery & Food IPC: Flour Protein & Ash Variability Control | Target Dough Temperature Control | Dough Absorption Control | Bake Profile Verification | Moisture Loss & Bake Yield Testing | Crust Color Uniformity Testing | Texture Profile Analysis (Bakery Crumb Quality) | Allergen Changeover Verification (Bakery)