Ask any pharmaceutical quality leader whether their computerized systems are under control, and you will usually hear a confident yes. The LIMS is validated. The ERP is validated. The MES and the electronic batch record system are validated too. The binders are thick, the signatures are in place, and on paper the program looks airtight.
Yet the findings that sink inspections rarely come from those headline systems. They show up in the places nobody thought to classify as “computerized systems” in the first place. The real exposure lives in the blind spots: the instruments, spreadsheets, interfaces, and clock settings that never made it onto the risk register.
This article walks through six categories of risk that consistently slip through the cracks, and what a mature data-integrity program does to close them.
1. “It’s just an instrument, not a system”
Walk into almost any QC lab and you will hear the same disclaimer during an inspection: that’s just a balance, or it’s only a pH meter. The thinking is that a piece of hardware sitting on a bench cannot possibly fall under computerized system management.
But modern lab instruments are rarely dumb. A balance, a pH meter, a conductivity meter, or an older spectrophotometer often carries embedded software that stores readings, runs calculations, and even supports basic user accounts. The moment that device processes, calculates, or stores data used for a GMP decision, it belongs in the scope of computerized system control, with the same expectations for access control, data protection, and lifecycle management as any server-based application.
The cracks are predictable. These devices frequently run on a single shared administrator login, so there is no individual accountability. Operators can delete a weighment and re-weigh without leaving a trace. Firmware versions are never assessed for whether they need validation. And the clock can be changed by whoever happens to be standing there.
A balance with memory is, functionally, a recorded-weighing system. If an analyst can wipe one result and redo it with no footprint, that is no different from a paper record someone can quietly tear out and rewrite. The instrument shell just makes it easy to look the other way.
2. Spreadsheets: the largest gray zone in the compliance program
Excel is everywhere in a pharmaceutical operation. Stability trends live in spreadsheets. Environmental monitoring summaries get calculated in them. In some sites, the key figures inside an OOS investigation were produced by a spreadsheet too.
The tool is not the problem. The loss of control is. A proper LIMS enforces role-based access, an audit trail, and version control. A spreadsheet often lives on one person’s desktop, can be opened and edited by anyone, and lets formulas get overwritten by accident, with no record of whether the file changed in the last six months.
The most dangerous pattern is the silent handoff: data leaves an instrument, passes through several spreadsheet transformations, and only then lands in a report. Each transformation is a black box outside the audit trail. An inspector who does not press on how that number was derived will almost never surface the risk, which is exactly why organizations miss it themselves.
Any spreadsheet that feeds a GMP decision should be assessed for whether it needs to enter the computerized system program. Version control, locked formulas, restricted access, and a change log are not optional extras. A validation framework built on top of an uncontrolled calculation tool is a framework with a hollow floor.
3. The data gaps between systems
A single application can be locked down tightly and still leak risk at its edges. Think about the journey of a result: an instrument sends data to the LIMS, the LIMS passes a disposition to the ERP for batch release, environmental sensors feed a SCADA layer that rolls up into a reporting system. Every time data crosses from one system to another, it can be lost, altered, or silently mismatched.
Most companies validate each system in isolation. Far fewer validate the connection itself. When a transfer fails, does the system raise an alert? Can data get truncated or mangled during a format conversion? When someone manually re-keys data between two platforms, is there a second check?
These seam risks are especially sneaky because, when something breaks, both systems may show a clean log that says “everything normal.” The actual problem sits in the space between them, the gap that nobody owns.
4. “Data orphans” left behind by retired systems
Every system has a lifecycle. An old LIMS gets replaced. A legacy MES is decommissioned for a new platform. That is ordinary technology turnover. What is not ordinary is how little thought goes into the handoff.
The questions tend to surface too late. Do the old records need to move to the new system? Can you carry over the audit trail, the metadata, and the history, not just the final numbers? If you do not migrate, will the retired system still open and read correctly inside the data retention period? And when the one engineer who knew how to boot that discontinued platform leaves, what happens the day an investigation needs to pull a historical record?
These data orphans stay hidden until the moment you need to trace something, and by then it is usually too late. A decommissioning plan deserves the same rigor as a go-live plan: explicit rules for migration, long-term retention, and proof that the archived data can still be read.
5. The time lag in access control
Nearly every company has a user-access policy. The gap is between the policy and the reality of execution. An employee leaves and IT disables the account two days later. A transferred analyst keeps an old high-privilege role. A temporary admin login borrowed for a project never gets switched off.
Individually, each looks like a small slip. Together they are the most direct route to a data-integrity breach: someone who should not have the ability to change critical records keeps that ability anyway.
The easier blind spot is vendor remote access. Equipment makers often keep standing backdoor credentials for maintenance, and the site frequently cannot say what the vendor can actually see, what they can change, or when they last connected. Cloud-based SaaS makes this worse. Where the data physically sits and who can touch the database directly from the backend is often genuinely unclear to the customer.
Access governance is not about what the policy document claims. It is about whether current permissions match each role’s actual need, verified on a recurring schedule rather than after an incident.
6. System time: the overlooked foundation everything rests on
An audit trail is only as trustworthy as the timestamps behind it. That sounds obvious, and it is also the piece most teams never check.
Across a site, devices on different floors and different production lines often drift. One clock runs ten minutes fast. Another lags half an hour. Some can be reset by an ordinary user on a whim.
When system time can be manipulated, the chronological order of the audit trail stops meaning anything. In theory, a record could be edited “back” to a previous point in time with no anomaly anyone would notice. This is the most basic slab in the data-integrity foundation, and the one most often left unexamined.
Where the risk really hides
These six share one trait: none of them sit inside the systems a company proudly invested in. Each one got excluded from formal management precisely because it seemed unimportant: just a small tool, temporary, about to be retired. That instinct is exactly what lets the gaps persist.
The hard part of computerized system risk management is not governing the heavily scrutinized platforms. It is finding the corners that generate, move, and store quality data while never quite entering the field of view. A mature quality system should keep asking one question long after the validators have gone home: what else is producing GMP data right now that we have never actually looked at?
The answer is usually hiding behind the most casual sentence in the room: “that doesn’t really count as a system, does it?”

Comments are closed