What is Process Development and when do you need it?
Process development is the bench-to-batch work that takes a molecule someone made once, at small scale, and turns it into a process you can run the same way, at a useful scale, with results you can predict. In a discovery or research lab a protein gets expressed in a few shake flasks and purified by hand, or an API is made on a route that works for grams. None of that survives contact with a clinical batch. Process development is where a CDMO defines the actual unit operations, sets the operating ranges, and proves the process gives you the yield, purity, and quality you can defend later.
You typically need it once you have a locked candidate and you are heading toward GMP material for the clinic, so it sits squarely in the CMC stage, usually feeding the IND-enabling and Phase 1 supply effort. For a biologic the work splits into upstream (cell culture or fermentation: media and feed strategy, bioreactor conditions, titer) and downstream (harvest, capture chromatography, polishing steps, viral clearance, ultrafiltration and diafiltration, final formulation). For a small-molecule API it is route selection and optimization, reaction conditions, crystallization and polymorph control, and impurity fate-and-purge. Different modality, same goal: a process that is reliable, scalable, and characterized well enough that the next batch is not a surprise.
The reason teams treat this stage seriously is that everything downstream inherits it. Your analytical methods, your control strategy, your stability program, and eventually your Module 3 filing all rest on the process you define here. Cut corners in process development and you pay for it in failed engineering runs, a tech transfer that does not reproduce, or a comparability headache when you scale for Phase 3. Get it right and the GMP batches that follow are far less likely to slip your clinical timeline.
What does a Process Development CDMO actually do?
A process development group does the experimental work that converts a candidate into a manufacturable process and the documentation behind it. The exact menu depends on modality, but the shape is consistent: optimize the steps, understand which parameters matter, lock a control strategy, and transfer the process cleanly to the manufacturing floor. Most of this can run non-GMP, since the point is to define and de-risk the process before you commit to expensive GMP runs.
What you are buying is process knowledge, not just hands. A strong group uses design of experiments (DoE) to map how critical process parameters drive critical quality attributes, runs scale-down models so a 2 to 5 liter bench bioreactor predicts behavior at 200 or 2000 liters, and builds the process characterization that supports your eventual control strategy and process validation. For an API, that same discipline shows up as reaction kinetics, impurity tracking, crystallization and polymorph studies, and a route that is safe and economical at scale.
- Upstream development (biologics): cell line evaluation and fit, media and feed optimization, bioreactor scale-up, titer and productivity improvement, and harvest clarification.
- Downstream development (biologics): chromatography capture and polish, viral inactivation and filtration, UF/DF, and aggregate and impurity removal to hit your purity target.
- Process synthesis and optimization (small molecule): route scouting and selection, reaction optimization, crystallization and polymorph control, and impurity fate-and-purge studies.
- Process characterization and DoE: identifying critical process parameters and quality attributes, defining proven acceptable ranges, and building the control strategy.
- Scale-up and scale-down models: translating a bench process to pilot and engineering scale, with a representative scale-down model for later characterization.
- Formulation and developability support, coordinated with analytical development so you can measure what you make.
- Tech transfer: protocols, batch records, risk assessments, and on-floor support to move the process into engineering and GMP runs without losing yield or quality.
How to choose a Process Development CDMO?
The first filter is modality fit, and it is rarely close. A group that develops CHO-based monoclonal antibody processes every week is not the right home for an AAV gene therapy, an mRNA and LNP product, an autologous cell therapy, or a small-molecule API, and the reverse holds too. The unit operations, the scale-down models, and the analytical baggage differ completely. Match the CDMO to a process they run routinely, not one they are stretching to win.
After modality, the decision is mostly about whether the work will survive scale and transfer. The reason process development matters is that it feeds GMP manufacturing, so the partner that does it well is often the one that can also make your clinical material or hand off cleanly to the site that will. Ask how their process development connects to their GMP floor, how they document and de-risk tech transfer, and whether their scale-down model genuinely predicts the scale you need. Use this checklist when you compare suppliers:
- Quality and GxP status: confirm where the non-GMP development work ends and GMP begins, and that the same supplier (or a clean transfer) carries you across that line. Process development itself is usually non-GMP, but it has to set up GMP-ready manufacturing.
- Capacity and lead time: real availability and a realistic slot, plus turnaround on development runs. A great group booked solid for months can be slower than a good one with an open bench.
- Modality and indication fit: routine experience with your specific process (mAb, viral vector, mRNA/LNP, cell therapy, peptide, ADC, or small-molecule API), with relevant case studies, not a general capabilities deck.
- Region and regulatory track record: where the site sits, which agencies (FDA, EMA, PMDA) have inspected the connected GMP operation, and whether their process work has supported filings before.
- Data quality: clean DoE design, traceable process data, sound scale-down models, and honest reporting of runs that failed, because the failures tell you more than the wins.
- IP and confidentiality: clear ownership of process improvements and know-how, defined data and material transfer, and confidentiality terms you can live with on a process that is part of your asset's value.