Skip to content
Use cases

Veterinary vaccines

Evidence that every batch ran the same way

A veterinary vaccine manufacturer with a broad portfolio and few batches per product wants a quantitative record of process consistency, built from data it already records.

The challenge

In a bacterin the antigen is made in days and judged weeks later. Between the fermenter and the potency verdict sit inactivation, formulation and testing, and nothing in between says whether the antigen was good. A weak component is carried into every dose of a blend.

Statistical confidence in a process comes from repetition, and with a wide portfolio no single product accumulates runs quickly. When a batch is weak, it is hard to say whether it was unusual or whether the process simply varies that much.

The client

A European manufacturer of veterinary immunologicals — bacterial, viral and autogenous vaccines — running many different cultures in modest numbers on shared equipment.

Our approach

We model the equipment, not only the product: a reference model of each fermenter, built from the runs of every product it hosts. During a culture the twin:

  • compares the live run against the pattern of runs that passed
  • names the variable that deviates
  • scores each finished batch, building a consistency record that accumulates across products

Estimated effects

1 model per vessel

shared by every product that runs in it

Evidence is pooled across products, which matters when each product has few runs.

At the fermenter

instead of at potency testing

A weak batch is spotted before it consumes inactivation, formulation and weeks of testing.

0

changes to a registered process

The model is process monitoring on data already recorded; its regulatory use is a decision for the manufacturer’s QA and regulatory affairs.

We do not promise a percentage. How much of the batch-to-batch variation is visible in the recorded signals is exactly what Phase 0 measures, on the manufacturer’s own historical runs. A process model does not replace a potency test.

How the PoC runs

  1. 01

    Phase 0 · 4 weeks

    Data assessment; choice of vessel and products.

  2. 02

    Data buffer · 2–4 weeks

    Cleaning, digitising and time-aligning the culture data.

  3. 03

    Phase 1 · 8–12 weeks

    The reference model, tested on held-out batches.

  4. 04

    Next step

    Extending the consistency record to further vessels and products.

What we asked for

  • Vessel description

    A drawing or P&ID of one fermenter, with measurement points and ranges.

  • Culture trends

    Temperature, pH, redox, agitation, gassing and dosing as recorded during the culture.

  • Batch outcomes

    Antigen content, potency result and release verdict per batch, however they are expressed.

  • Batch history

    Batch number, product, strain, start and end, medium lot and recorded deviations.

What you get

  • Earlier detection

    A weak batch spotted at the fermenter.

  • Documented consistency

    The evidence base that a move from animal to in-vitro potency testing rests on.

  • Pooled evidence

    Across products, even when each has few runs.

  • Autogenous batches

    Useful where volumes are smallest and turnaround matters most.

Data can come in any format — chart-recorder printouts, spreadsheets, PDF, scanned batch records. The work runs on a copy of historical data, outside GxP and without touching production. No new sensors, IT integration, remote access or change to a registered process are needed to start.

Have a similar process?

Tell us about it. We’ll tell you where a model pays off first and what data you’d need.

Book a technical call

Or apply for an ADT pilot →