For years, “continuous manufacturing” has been one of biopharma’s favorite phrases — invoked in conference keynotes, regulatory guidance, and strategy decks as the obvious next step. On the upstream side, perfusion culture has made it real for a while. Downstream — the purification half of the process, where I spend my days — has been slower to follow, and for reasons that are worth understanding rather than glossing over.
What “continuous” actually means downstream
In a traditional batch process, each purification step runs to completion before the material moves on: load a chromatography column, wash, elute, collect, hold, then start the next unit operation. It’s robust and well understood, but it’s also full of waiting — tanks holding intermediates, resins sitting idle between cycles, equipment sized for peak volumes it only sees briefly.
Continuous downstream processing connects those steps so material flows through the train more or less without stopping. Multi-column chromatography lets one operation run steadily by cycling several smaller columns out of phase. Connected polishing and filtration steps hand material off directly rather than parking it in a tank. Done well, the result is a smaller footprint, better resin and buffer utilization, and — this is the part I find most compelling — more consistent product, because the process spends more time at steady state and less time in transient start-stop conditions.
Why it’s harder than the slides suggest
Here’s the honest part. Connecting unit operations doesn’t just save tank space — it removes the buffers, in both senses of the word, that batch processing quietly relied on. In a batch train, a hold tank isn’t only storage; it’s a shock absorber. If one step runs a little long or a little off-target, the next step doesn’t care, because it starts fresh from a well-characterized pool. Take the tanks out, and you couple the steps together: variability in one operation propagates directly into the next.
That changes what “process control” has to mean. You need to see the process in real time — which is why continuous downstream and process analytical technology (PAT) are really the same conversation. You need control strategies that respond to what the sensors say, not just recipes that assume nominal conditions. And you need a much clearer picture of how disturbances travel through a connected train, because the margin for “we’ll fix it in the next step” is gone.
This is also where the modeling side gets interesting. When you can’t run a full factorial DOE on every connected configuration, mechanistic and hybrid models — and increasingly, machine-learning-augmented approaches — become the way you explore the design space without burning through material and calendar time. A digital twin of a purification train isn’t a gadget; it’s how you reason about a system that’s genuinely harder to intuit than a sequence of independent steps.
Where this actually lands
I don’t think the future is “everything continuous, everywhere.” I think it’s selective — teams intensifying the steps where the payoff is real, running hybrid batch/continuous trains, and being honest about where the added control burden isn’t worth it yet. The interesting work isn’t cheerleading the transition; it’s the engineering judgment about which steps, which molecules, and what has to be true about your process understanding before connecting things is a good idea rather than a fragile one.
That judgment — part separation science, part process control, part modeling — is what I find most worth writing about, and it’s what I’ll keep digging into here.
These views here are my own. I’d genuinely like to hear how others are approaching the batch-to-continuous question — reach me via the Let’s Connect page.]