OEE on a pharma line — better performance without breaking GMP validation.

OEE in pharma only competes with GMP when you try to win by stepping outside the validated range. iLEAN measures OEE per batch, per shift and per line, breaks down the micro-stops the SCADA never saw, and proposes optimizations inside the approved range. The line performs better without a single comma of the validation changing.

← See all pharma solutions

Pharma blister line with an iLEAN Edge camera counting micro-stops and an operator checking OEE per batch on a GMP tablet — iLEAN pharma OEE per batch
The problem

The OEE you see on the plant dashboard is not the real OEE — it is the one that fits inside the SCADA.

The OEE metric in pharma leaks in two places at once:

  1. The wrong granularity. The SCADA gives OEE per shift and per line — not per batch. In a plant where the line runs two products in the same shift, mixing the losses of both hides the root cause: the changeover between A and B disappears inside an aggregated metric.
  2. Invisible micro-stops. Short blister jams, an empty feeder, operator touch-ups, false rejector alarms — none of it reaches the SCADA, or it arrives aggregated over a long period. The "performance losses" column in the monthly report exists, but it is painted with data that does not represent what really happens on the line.

On top of that, any optimization proposal hits the validation wall: changing a process parameter to raise speed means opening change control, revalidating and waiting months. The result: the continuous improvement plan fizzles out, and the plant gets used to an OEE 10 or 15 points below what is possible, because nobody wants to touch what is validated. The system works — and leaves capacity on the table.

How it fits the IRIS system

iLEAN does not touch your validation — it opens the lever that was already inside the approved range.

OEE in pharma does not need more sensors on the line: it needs to cross-reference what is already there (SCADA, validated MES, ERP planning, the operator's notebook) with what is lost right now (micro-stops nobody assigns, changeovers with no target time, operator touch-ups). iLEAN acts as the putty that seals that crack without asking you to touch the validated MES or the process dossier.

Edge counts what the SCADA aggregated and lost. Brain breaks the losses down per batch. The agent proposes optimizations inside the approved range and the person validates — outside the range, it opens change control. The validation does not change.

The three iLEAN pieces applied to pharma line OEE:

  • iLEAN Edge — a terminal with machine vision (CNN) over the blister/tablet/injectable line. It counts every micro-stop, assigns it to the right reason (jam, feeder, touch-up, false rejector alarm) and records the reject visually. It works with no network: if the plant loses WiFi, Edge keeps counting. Inference happens in the box, not in the cloud — the process data never leaves the OT ring.
  • iLEAN Brain (Central + Agents) — receives the Edge data, cross-references it with the validated MES (production order, recipe, batch events) and with the LIMS (quality results), and breaks OEE down per batch/shift/line/product. Losses stop being a gray column and become an actionable list.
  • iLEAN Agent — proposes controlled optimizations: adjustments that fall inside the validated range are proposed to the supervisor to apply immediately; adjustments that fall outside are packaged as a change control proposal with the evidence already prepared. QA decides. The validation is never altered behind anyone's back.

See the full IRIS architecture →

Before and after

Classic pharma OEE vs. OEE per batch with iLEAN

AspectClassic OEE (SCADA + monthly report)With iLEAN Edge + Brain + Agent
GranularityPer shift and per line (aggregated)Per batch, shift, line and product
Micro-stopsLumped into an "other losses" blockCounted one by one with a reason assigned
ChangeoverTarget time in Excel, never checked against realityTime measured per changeover, compared with the target
Improvement proposalMonthly, in a meeting, with old dataContinuous, with real data, inside the validated range
GMP complianceAny change triggers months of revalidationOnly what falls outside the range enters change control
Traceability of the improvement"You used to run at 58 and now you run at 64" — no breakdownPart 11 audit trail of the before/after per lever
Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

  • GMP plant with 2-4 blister, tablet or injectable lines, current OEE between 45% and 65% (the typical range for the sector), a validated vertical MES and a standard SCADA.
  • First value expected within a few weeks: the correct breakdown of micro-stops and two or three concrete levers inside the validated range.
  • Expected reduction of unplanned losses: ≥30% in the first 6 months, without touching the validation dossier.
  • Indicative payback between 4 and 9 months, dominated by recovered capacity (a third shift avoided, extra batches within the same plan).
  • Sector benchmark: +40% productivity in the most digitalized sectors compared with the least digitalized ones [1]. The gap is widening; whoever does not break the losses down is already losing ground.

And the technical director's reasonable doubt

“What if the AI proposes an optimization outside the validated range and compromises product integrity?” — the rule is strict: whatever falls inside the approved range is proposed to the supervisor; whatever falls outside is packaged as change control with evidence. The AI does not act on the line without a person signing. And even so, the reliability of the analysis is high: in anchored tasks (reading a value from the SCADA and comparing it with the range in the dossier) the best models brought error below 1.5% [2]. The three safety rings exist precisely for this.

[1] +40% productivity in the most digitalized sectors (2000-2021). Source: Fundación BBVA/Ivie.

[2] OpenAI paper “Why Language Models Hallucinate”, 2025 — reliability of AI in anchored tasks.

Frequently asked questions

What people ask about OEE per batch on a pharma line with AI

How does measuring OEE fit with staying GMP compliant?

OEE (Overall Equipment Effectiveness) and GMP do not compete — they compete when you try to improve OEE by stepping outside the validated range. iLEAN measures the three components (availability, performance, quality) of every batch continuously, but it only proposes optimizations inside the range approved by the validation dossier. Anything outside that range is never applied on its own — it is flagged as a change control proposal and the QA person decides. The cumulative OEE improvement is real, and the validation stays exactly as it was.

Is it measured per shift and per line, or only plant-wide?

Per batch, shift, line and product, all four at once. Batch-level granularity is what makes the metric useful in pharma: the blister line runs two batches of different products in the same shift and the OEE of each one is broken down separately, because the root cause of the loss may sit in the changeover between them. Brain cross-references the metric with the production order, the recipe and the micro-stops Edge captured, and breaks the losses down as far as the data allows.

Does it work with an already validated MES?

Yes. iLEAN does not replace your validated MES: it reads from it what is already right (production order, master recipe, batch events) and fills the cracks with Edge (micro-stops the SCADA aggregates and loses) and Tracer (reading the panel of the old cartoner that exposes nothing to the MES). iLEAN's GAMP 5 validation package is handed to the customer's QA team; the MES remains the MES. The customer keeps the validation they already have.

Does it detect the micro-stops the SCADA does not see?

Yes — that is exactly the crack where iLEAN Edge adds value. Short stops (a blister jam, an empty feeder, an operator touch-up) usually never reach the SCADA, or arrive aggregated over a long period, which is why they disappear from the OEE metric while still weighing on real performance. Edge with vision over the line detects them one by one and assigns each to the right reason. For the first time, the production manager sees the true color of the 'performance losses' column.

What OEE improvement should you expect?

Estimate to be validated with your numbers: on a pharma line with current OEE between 45% and 65% (the typical range in standard blister/tablet/injectable) the first value with iLEAN shows up within a few weeks — usually as the correct breakdown of micro-stops, which opens 2 or 3 concrete levers. Expected OEE improvement: ≥30% reduction of unplanned losses in the first 6 months, inside the validated process range. Indicative payback between 4 and 9 months, dominated by recovered capacity (a third shift avoided, extra batches that fit into the same plan).

Let's talk

Tell us your case and in 48h we'll send you the estimated ROI of your OEE per batch.

We work on your pharma line's real data, not ours. Diagnostic with no commitment.

Request estimated ROI in 48h See pharma