A lot of a pharmaceutical plant was bought before “connectivity” was a purchasing requirement. A bioreactor, a lyophilizer, a filling line — installed, qualified, and running well, with years of validated service still ahead of it. The machine works. What it won’t do is easily hand over its data. The useful question is not “when do we replace this?” but “how do we read what it’s already telling us — without disturbing a line that’s validated and in production?”
Replacing the machine is the expensive answer
The reflex — rip out the old equipment and buy something that speaks a modern protocol — is usually the wrong first move in pharma. New equipment means new capital, new installation, and, critically, requalification: the machine has to be re-validated in its process before it can make product again. That is months of work and lost capacity for a line that was, until you touched it, producing perfectly good batches.
The equipment isn’t the problem. The problem is that the data — cycle counts, temperatures, alarms, run states — is trapped inside a controller that was never asked to share it. A retrofit solves that directly: it adds a way to read the data off the existing machine and leaves the machine itself alone.
The retrofit toolkit
The good news is that the components for reading data off industrial equipment are mature, standardised, and mostly vendor-neutral. You don’t need much.
- OPC UA (OPC Unified Architecture). The vendor-neutral standard for industrial data. It gives every value a consistent name and structure, so a temperature from a 15-year-old machine looks the same to your platform as one from new equipment. Many controllers can expose an OPC UA server directly; older ones get one added alongside.
- MQTT. A lightweight publish-subscribe protocol that moves small messages from many machines to a broker with very little overhead. It suits patchy or constrained networks and pairs naturally with OPC UA for the last hop to the cloud or data platform.
- Edge gateway. A small industrial computer that sits next to the machine, speaks whatever the old controller speaks (Modbus, a serial protocol, raw PLC tags), and re-publishes it cleanly as OPC UA or MQTT. It's the bridge between the machine and the platform — and the place where you buffer, filter and timestamp data at the source.
- PLC and SCADA. The programmable logic controller (PLC) runs the machine; the supervisory control and data acquisition (SCADA) system watches over it. A retrofit taps into these rather than replacing them — reading tags the controller already computes.
For modular additions — a new skid, a bolt-on unit — the Module Type Package (MTP) standard is worth knowing. MTP describes a piece of equipment in a portable, vendor-neutral file so it can be integrated into a plant with far less manual engineering — the industrial equivalent of plug-and-produce. It won’t retrofit a machine that predates it, but for anything new you add alongside the old line, it keeps the integration effort low.
The IT/OT boundary — and why it matters
The moment you connect a production machine to anything, you meet the IT/OT boundary: the line between information technology (your networks, servers, cloud) and operational technology (the controllers and equipment that physically make product). These two worlds have different priorities. IT optimises for data flow and patch cycles; OT optimises for uptime and safety, and its equipment often can’t be patched at all without requalification.
Cross that boundary carelessly and you expose validated, safety-critical equipment to the wider network — the exact risk that keeps plant managers from saying yes to digitalization. Cross it deliberately and it’s routine:
- Segment the networks. The OT side stays on its own isolated network. The edge gateway is the single, controlled point where data moves across — not a flat network where IT and machines share the same wire.
- Go one-way where you can. For pure monitoring, data only needs to travel out of the OT zone. A one-directional path (up to a data diode for the strictest cases) means nothing on the IT side can reach in and change the machine.
- Read, don’t control. The lowest-risk retrofit takes data off the PLC without ever writing to it.
You can take data off a controller without taking over the controller. That one distinction is what makes a retrofit safe to run on a validated line.
Taking over control — closed-loop automation, remote setpoints — is a real and valuable step, but it’s a later one, with its own validation. Reading first gets you the visibility, builds trust, and disturbs nothing.
Keeping it compliant and validated
In a GxP environment the connectivity is only acceptable if it respects the rules the plant already runs under. A well-designed retrofit is built to leave the validated state intact.
- Don't disturb the validated line. A read-only overlay changes nothing about how the machine makes product, so the equipment's existing qualification stands. The retrofit itself is validated as an add-on, not as a change to the process.
- Data integrity — ALCOA+. Data has to be Attributable, Legible, Contemporaneous, Original and Accurate (plus Complete, Consistent, Enduring, Available). Capturing at the edge, timestamped at the source, with no manual re-keying, is how you meet that.
- Audit trail. Who saw what, when, and whether any value was changed — recorded and tamper-evident. For a read-only retrofit this is mostly about provenance: proving the data is a faithful copy of what the machine produced.
- GAMP 5 and Annex 11. GAMP 5 gives the risk-based framework for validating the computerised system; EU GMP Annex 11 sets the expectations for computerised systems in GMP. A retrofit's software layer is scoped and validated against both.
What the data unlocks
Getting the data off the machine is the enabling step. The value shows up in what you can finally do once it’s flowing:
- Real-time visibility and OEE. Overall Equipment Effectiveness (OEE) — how much good product a machine makes versus its theoretical maximum — stops being a monthly spreadsheet estimate and becomes a live number you can act on.
- Predictive maintenance. Trends in vibration, temperature or cycle time warn you before a machine fails, instead of after — turning unplanned downtime into scheduled work.
- A data platform and digital twin. Standardised, trustworthy data from the floor is the raw material for a plant-wide data platform and, eventually, a digital twin: a live model you can use to test changes and optimise the process without touching the real line.
None of that requires new machines. It requires the data those machines already generate — read off them safely and delivered somewhere it can be used.
Where this leads
A retrofit is deliberately small in scope: read the data, cross the IT/OT boundary safely, keep the validated line untouched, stay compliant. But it’s the first move in a much larger programme. Once the floor data is standardised and trustworthy, it feeds the monitoring, maintenance and modelling that a modern manufacturing operation runs on. That connect-the-legacy-equipment step is exactly where our Digital CDMO work starts — reading what your existing plant is already telling you, and building the data platform and digital twin on top of it, without asking you to replace a line that still has years of good service left.