The Three Types of Bad Data on the Pharma Floor: Wrong, Missing, Non-Actionable
Walk into a pharmaceutical manufacturing facility mid-shift, and the data problem is not immediately visible. The MES is running. The DCS is logging process parameters. The EBR is open at a terminal somewhere on the floor. From the outside, the data infrastructure looks functional: data is being generated, stored, and transmitted.
What the data trail often does not tell you is that a significant portion of it is unreliable, because of the conditions under which operators capture it, enter it, and act on it. Those conditions produce three distinct categories of failure:
- Wrong data
- Missing data
- Data that exists but cannot drive a decision
These three are not the same problem. They do not share a root cause. They do not show up the same way in an inspection. They are not fixed by the same intervention. Most data integrity programmes focus on one of them, usually the compliance visibility of the first, and leave the other two largely unaddressed.
The Data Integrity Conversation Usually Starts in the Wrong Place
The regulatory framing around data integrity centres on records after the fact. ALCOA+, FDA 21 CFR Part 11, EU GMP Annex 11, and the ISPE GAMP guidance on records and data integrity all focus on attributability, accuracy, and the completeness of a record once it exists. Those are the right things to care about from an inspection standpoint. They describe what auditors look for when reviewing records, not what is happening on the floor when the data is created.
That distinction matters, because data integrity failures start at the point of work, in the gap between where the process is happening and where the data about it gets entered. Regulators know this. The FDA's 2024 warning letter record shows missing records, incomplete documentation, and inaccurate data as persistent top citations. More than half of all FDA citations in FY2024 related to inadequate procedures. Named cases bear it out: the December 2024 warning letter to Indoco Remedies cited audit-trail review controls that left gaps in how data was checked. Much of this traces back to the physical conditions under which documentation happens.
The three types of bad data at a glance
Type 1: Wrong Data
Wrong data is the most legible category, because it is the one that shows up directly in regulatory citations. The FDA's review of 2024 inspection data identified data transformation errors, unverified manual calculations, and accuracy failures as recurring themes.
Some of this is deliberate falsification, and those cases carry the most serious language in the warning letters. The more common source of wrong data on the manufacturing floor is documentation completed retrospectively: from memory, under time pressure, or in physical conditions that increase transcription error.
When an operator is managing a process at one end of a fill-finish suite and the MES terminal is fixed ten metres away, the documentation does not happen at the moment of the event. It happens later, after the operator has walked to the terminal and their attention has shifted. The ALCOA+ requirement for contemporaneous records exists precisely because of this gap. Data entered from memory is structurally more likely to be wrong than data entered at the moment of observation. The physical distance between the process and the documentation point produces inaccurate records with consistent predictability.
Type 2: Missing Data
The more common form of missing data is the entry that was intended but never made. A parameter that should have been logged at 14:30 gets entered at 16:00 without a timestamp. A verification step deferred during a busy period is then forgotten entirely.
FDA inspectors flagged this pattern explicitly across 2024 enforcement actions. Incomplete records, with missing attribution, timestamps, or ALCOA+ attributes, appeared consistently across warning letters issued to manufacturers in multiple geographies and facility types. The ISPE GAMP framework classifies GxP data by the decision it supports. Data that is absent from the record cannot support any decision, regulated or otherwise.
The mechanisms that produce missing data are primarily structural. Operators managing multiple process steps across a large floor, working within shift patterns that create hand-off pressure, and reaching documentation through terminals that require them to leave their station, will generate documentation gaps.
Type 3: Non-Actionable Data
The third category is the one data integrity programmes most frequently overlook, because non-actionable data is not a compliance failure in the conventional sense. The data exists and passes an audit trail review. It still cannot drive an operational decision, because by the time it reaches the person who needs to act on it, the window for action has already closed.
This is the pattern in many MES and DCS deployments in larger manufacturing environments. Process parameters are logged continuously. If the operator responsible for that line is not positioned to receive alerts in real time, the alert is effectively non-actionable.
The ISPE GAMP guidance makes an important distinction here. It classifies data as regulated (used for regulated decisions), operational (used for business process decisions), or unnecessary, meaning data that adds no context, metadata, or meaning to the activity.
Non-actionable data occupies an uncomfortable position. It is often regulated in classification and captured as part of a validated system, yet operationally indistinguishable from data that was never captured at all.
The question regulators ask is whether the data exists. The question operations teams should be asking is whether the data is positioned to make a difference before the deviation has already occurred.
What It Looks Like on the Floor
Consider a packaging environment with a fixed MES terminal at one end of the suite. Operators work the line throughout the shift, logging parameters when they can get to the terminal. During high-output periods, some entries are deferred and completed from memory at the end of the run. An alert fires midway through the shift when a parameter drifts. It appears in the system log. The operator is at the other end of the suite and does not see it for eleven minutes. By the time the response is initiated, the process has continued outside specification for most of that window.
In this one scenario, all three types of bad data have been produced:
- Wrong data, from the retrospective memory-based entries
- Missing data, from the steps that were deferred and not completed
- Non-actionable data, from the alert that existed in the system but arrived too late to drive a response
Now consider the same environment with a mobile workstation at the operator's side, moved to the point of work. Entries can be made in real time. Alerts are visible immediately. The record reflects what happened, at the moment it happened, by the operator who was present. The data quality difference comes from one change: where the documentation interface sits in relation to the work.
Kinetic ID's ID-Flow 5 and ID-Flow 6 are designed for this deployment model. They are mobile, GMP-rated workstations that bring MES, EBR, DCS, and SCADA access to Grade C/D and Grade B/C environments, without requiring operators to leave the process to reach the system. The ID-View extends the same principle to fixed Grade B and critical-zone positions, where in-room access is required but limited floor space makes a fixed unit the appropriate solution.
The argument for point-of-work infrastructure in regulated manufacturing is often framed as a productivity case. The data integrity case is more direct. If the documentation interface is not at the point of work, the conditions for all three categories of bad data are built into the floor design by default.
What point-of-work capture is worth
Kinetic ID's GMP-rated ID-Flow workstations run in regulated environments with figures customers report back to us: up to 20% higher uptime across units, 9% saved in daily operator activity, and up to 90% lower overnight energy use on the Eco-Charger system. Fewer retrospective entries and faster alert response show up in deviation and audit data across a full shift. Our solutions team can walk you through the numbers for your environment.
Speak with a solutions consultant
The Questions Ops, QA, and Automation Teams Should Be Asking
For operations and automation engineers, the practical audit is not of the data itself but of the conditions under which it is created:
- Which terminals require operators to leave a process station to reach them?
- Which alert pathways assume an operator is within line of sight of a display?
- Which documentation steps are routinely completed outside the window in which the event occurred?
Those questions locate the structural sources of bad data before they surface as citation findings.
For QA leads, the data quality conversation is worth extending beyond ALCOA+ attribute completeness into where, specifically, documentation gaps cluster. If certain lines, workstations, or shifts generate a disproportionate share of incomplete or retrospective entries, that pattern is informative. It almost always maps to a point-of-work visibility gap.
For IT/OT integration leads, the question is whether the system access model assumes a fixed-terminal deployment, and what the downstream data quality consequences of that assumption are. An MES that is technically capable of real-time documentation at the point of work, but physically deployed in a way that prevents it, is not delivering on its capability. The IT/OT integration layer is only as good as the physical access points it reaches.
Data Quality Is an Infrastructure Problem Before It Is a Compliance Problem
The instinct when data integrity issues appear in warning letters or internal audits is to reach for procedural remediation: updated SOPs, additional training, closer supervisory review of audit trails. Those interventions address the symptom where it is visible. They do not address the structural conditions that produce wrong, missing, and non-actionable data across every shift.
The common thread across all three failure types is distance: between the operator and the documentation interface, between the alert and the person who needs to act on it, and between the event and the record of it.
Reducing that distance is an infrastructure decision. It affects deviation rates, inspection readiness, and the reliability of the data that manufacturing and quality teams use to make decisions.
If your mobile workstation range doesn't reach the point of work, here's how we can help
Kinetic ID builds GMP-rated mobile and fixed workstations for life sciences manufacturing, from Grade D packaging lines through to Grade B fill-finish suites. The ID-Flow range covers mobile deployments with hot-swap battery architecture and validated stainless steel construction; the ID-View covers fixed in-room access. Our solutions team can walk you through the options, grade requirements, and deployment configurations relevant to your environment.
Speak with a solutions consultant
Frequently Asked Questions
Calculate your productivity gain
Enter your site's operator headcount, shift pattern, and deviation rate, and get a payback estimate built on your numbers.
Try the ROI calculator
References
U.S. Food and Drug Administration (2024). Data Integrity and Compliance With Drug CGMP: Questions and Answers.
Eva Kelly, ERA Sciences (2025). These Were FDA's Top Citation Issues For Data Quality In 2024. Pharmaceutical Online.
ISPE GAMP. GAMP Guide: Records & Data Integrity.
https://ispe.org/publications/guidance-documents/gamp-records-pharmaceutical-data-integrity
ISPE GAMP. GAMP RDI Good Practice Guide: Data Integrity, Manufacturing Records.
ISPE GAMP. GAMP RDI Good Practice Guide: Data Integrity by Design.
https://ispe.org/publications/guidance-documents/gamp-rdi-good-practice-guide-data-integrity-design
U.S. Food and Drug Administration (2024). Warning Letter: Indoco Remedies Limited.
Scilife (2025). FDA Warning Letters 2025: Trends, violations, and how to avoid them.
https://www.scilife.io/blog/worst-fda-warning-letters-pharma
ALCOA+ Data Integrity Principles in Pharmaceutical Manufacturing. ifactoryapp.com.