The situation
A frozen ready meal manufacturer had a sauce component that passed every development trial and failed in the market. Returned product showed a watery layer separating from a grainy sauce, and the complaint rate rose with distance from the plant.
Why it is hard
The specification had been written for one freeze. Real distribution is never one freeze: the product is frozen at the plant, partly thawed during transfer, refrozen at the depot, held in a cabinet that cycles, and thawed again by the consumer.
A formulation tested once and used four times will pass in the laboratory and fail in the field.
Each cycle does the same thing. As water freezes it leaves the solution and concentrates everything still dissolved into a shrinking unfrozen phase. Polymer chains are pushed together and reassociate, which is retrogradation driven by freezing rather than by time. The structure changes a little each cycle, and the water it was holding is released.
What was done
The number of cycles the product actually meets was established first, by mapping the real distribution route rather than assuming the specification. The starch was then chosen for stabilisation against reassociation rather than for viscosity, since the failure is retrogradation and not thinning. A hydrocolloid was used alongside it to protect the system through the process, which the starch alone would not survive.
Testing was redesigned to run the real number of cycles rather than one.
What it shows
That the formulation was never the main failure. The test was, and it had been passing a product that the distribution chain was always going to break.
