Transfer the process.
Don't lose the reasoning.
A transfer package moves parameters and ranges. It almost never moves why those ranges exist, what was tried and rejected, or which deviations shaped them. The receiving site rediscovers all of it at their own cost, usually in the middle of the first engineering campaign, with the clock running.
What moves
The process, its ranges, and the evidence behind both
What the receiving site gets
A queryable history, not a 400-page PDF
What it is worth
Fewer surprise campaigns and a shorter comparability argument
The criterion, and the two it was chosen over.
Geometric similarity fixes D/T, so power per volume, tip speed and mixing time move against each other and no transfer holds all three. The platform computes each option against the receiving vessel, says which one the oxygen demand rules out, and leaves the choice with the engineer.
10 L → 2000 LStudy 1 confirmed condition, into the 2000 L train
day-6 shift · pH 7.00 · geometric similarity maintained
| 10 L bench | 2000 L production | |
|---|---|---|
| Working volume | 8 L | 1600 L |
| Tank diameter, T | 0.22 m | 1.29 m |
| Impeller diameter, D | 0.10 m | 0.58 m |
| D/T | 0.45 | 0.45 |
| Impeller | 3-blade segment | 3-blade segment |
| Power number, Np | 1.27 | 1.27 |
| Agitation | 100 rpm | to be set |
| P/V | 7.3 W/m³ | to be set |
| kLa | 14 h⁻¹ | ≥ 12 h⁻¹ required |
| Mixing time, θm | 12 s | to be set |
Three criteria, evaluated against the receiving vessel
P/V = Np ρ N³ D⁵ / V · v_tip = π N D · kLa = A (P/V)^0.5 (v_s)^0.5 · θm ∝ T^(2/3) (P/V)^−1/3
Constant P/V
viableholds power per unit volume
- N
- 31 rpm
- P/V
- 7.3 W/m³
- v_tip
- 0.95 m/s
- kLa
- 14.0 h⁻¹
- θm
- 39 s
Holds the power input, and with it the oxygen transfer.
Constant tip speed
failsholds impeller tip speed
- N
- 17 rpm
- P/V
- 1.2 W/m³
- v_tip
- 0.52 m/s
- kLa
- 5.7 h⁻¹
- θm
- 71 s
Holds shear at the impeller, and gives up most of the power.
Constant kLa
viableholds oxygen transfer, sparge raised
- N
- 27 rpm
- P/V
- 4.5 W/m³
- v_tip
- 0.81 m/s
- kLa
- 14.0 h⁻¹
- θm
- 46 s
Buys the transfer back through gas flow instead of agitation.
The three cannot be held together. Geometric similarity fixes D/T, so P/V, tip speed and mixing time move against each other, and the criterion is a choice about which one the process is most sensitive to. Oxygen demand at peak density needs 12 h⁻¹, which rules out constant tip speed at 5.7 h⁻¹.
Proposed basis: Constant P/V
Agitation
31 rpm
P/V
7.3 W/m³
kLa
14.0 h⁻¹
θm
39 s
Temperature, pH and feed carry across unchanged: they are process values, not scale-dependent ones. Only what the vessel changes has been recomputed.
What it costs, and what it does not
- Mixing time rises 12 s → 39 sBase and feed additions see a gradient before they disperse. Move the feed to a dip tube at the impeller and slow the bolus.
- pCO₂ accumulates at scaleHeadspace-to-volume falls with the cube root, so stripping is sparge-limited. Hold pCO₂ under 100 mmHg or the pH control fights it.
- Tip speed 0.52 → 0.95 m/sBelow the 1.5 m/s working limit for CHO, so no hydrodynamic change is expected.
- The kLa constant is not transferableBench uses a microsparger, the 2000 L a drilled-hole sparger. Bubble size and coalescence differ, so A is confirmed by a gassing study on the receiving vessel before the first cell.
Transfer package, drafted from the record
- Scale-up rationalethe criterion, the two it was chosen over, and why
- Parameter ranges for the 2000 L MBRcomputed, each carrying its correlation
- Gassing study protocolto confirm kLa on the receiving vessel
- Comparability planwhich attributes are compared against the bench runs
The cost you are already paying.
The package is a pile of documents, not a body of knowledge
Development reports, runsheets, batch records, CoAs and method files arrive as files. Answering a single question at the receiving site means somebody reading several of them and hoping they have the current version.
The people who know are the bottleneck
Every transfer runs on a handful of individuals from the sending site answering questions by email. Their availability, not the science, sets the pace, and when they move on, the answers go with them.
Comparability arguments are rebuilt from scratch
Demonstrating that the new site's batches match the old site's means assembling the same comparisons again, by hand, in a format QA will accept. It is the same work every time and it is never reusable.
The second CDMO costs as much as the first
Without shared definitions, every new partner site is a fresh integration and a fresh vocabulary. 'Feed rate' means something slightly different, and the analysis has to be redone rather than repeated.
On the platform.
The package becomes queryable
Every document is classified, segmented and linked to the molecule, study, batch or method it describes. A transfer question is answered in minutes, with the page it came from attached.
Rationale travels with the range
A parameter range carries the studies and runs that set it. The receiving site can see not just the number, but the argument, and challenge it on evidence rather than on instinct.
Comparable-batch analysis across sites
Runs from the sending site, the receiving site and your CDMO partners sit on one record with one set of definitions, so a comparison is a query rather than a project.
A shared ontology across the network
The same term means the same thing in every site's data. That is what makes the second site materially cheaper to bring on than the first.
Deviations carry forward as precedent
When the receiving site hits something the sending site already saw, the earlier investigation, its evidence and its conclusion are right there instead of buried in a QMS export.
The transfer document drafts itself
Tech transfer packages and campaign summaries generate from the record against your template, so the document that leaves matches the data that stayed.
Related capabilities
Bring us the work you already know the answer to.
We would rather you judge the platform against a result you can already check than sit through a scripted tour.
Book a demo