Scaling a Novel Food Product: Why Process Comes Before Equipment

The call usually goes like this: "We have a product nobody else makes. We've been doing it by hand. We need a machine." And the natural next step seems to be to design the machine.

It's the wrong next step, and it's an expensive one. A machine is a set of decisions about rates, temperatures, forces, timings, and tolerances. If you haven't proven what those numbers need to be, you're designing around guesses — and a production machine built on guesses is a very expensive prototype. Here's the sequence that works.

Step 1: Find the transformation that matters

Every novel product has one step that's actually novel. The rest — conveying, filling, packaging — is solved. A clear-ice producer's hard step was freezing at a controlled rate; moving and cutting ice is ordinary. A depositor for an unusual topping has a hard step in the deposit and an ordinary one in the conveyor.

Name that step. Everything else in the project should wait behind it, because if it can't be done, nothing else matters — and if it can, its parameters drive the design of everything around it.

Step 2: Prove it on the cheapest possible hardware

Build a rig, not a machine. A benchtop fixture, an instrumented tank, a hand-cranked version of the mechanism. The goal is to isolate the transformation and vary one thing at a time — temperature, rate, pressure, dwell — while measuring the result. Ugly is fine. Fast is the point.

On the clear-ice program, the first phase established that the product could be made at all, and revealed the thing nobody expected: off-the-shelf refrigeration controls produced clear ice with cracks. The variable that mattered wasn't how cold — it was how fast the temperature changed. That single finding reshaped the entire production system. No amount of machine design would have found it; only running the process did.

This phase should end with documented parameters and a demonstrated result — "at these settings, the product comes out right, repeatedly." Twenty-seven documented test runs, in that case. If you can't get there on the rig, you've learned that cheaply. If you can, you've bought the specification for the machine.

Step 3: Scale the transformation, not the whole line

The second phase takes the proven parameters to production dimensions — a full-size cold plate, a full-width deposit head — and asks whether the physics hold at scale. Often they mostly do, with new problems at the edges: heat transfer that worked in a small vessel fails in a large one; a mechanism that indexed cleanly at one per second jams at four. This is still R&D, still on purpose-built rigs, still cheap compared to a production machine. It ends with a measured cycle time and a measured yield at production scale.

Now — and only now — you know what the machine has to do. The design brief writes itself, and it's built on measured numbers, not hopes.

Step 4: Design for production, and guarantee it

With a proven process, the machine design becomes an engineering problem instead of a research problem. Rates, loads, and controls are specified from data. Custom controls, if the process needs them, have already been prototyped. And because the numbers are real, the builder can put a written performance guarantee on the production system — output per day, availability — which is something no honest builder can offer for an unproven process.

What this costs, and what it saves

Phased R&D on a novel process typically runs $40K to $150K over two to five months, depending on how novel. That sounds like a lot until you compare it to the alternative: a production machine designed on assumptions, at $300K to $2M, that has to be redesigned in the field. We've seen the field-redesign version — the machine that "works but needs constant tending," the line that needs more operators with the machine than without it. That's what skipping the proving phase buys.

The R&D phase also produces something valuable on its own: documented, proven process parameters. That's an asset. It's what you show an investor, what you hand a co-manufacturer, and what protects you if you ever need to build a second machine somewhere else.

Signs you're skipping steps

• You're getting quotes for a production machine and nobody's asked to see the product made.
• The builder's proposal has no proving phase and no numeric acceptance criteria.
• The critical step has never been run at more than a fraction of target rate.
• Someone has said "we'll figure that out during commissioning."

Each of those is a machine that will be redesigned on your floor.