DOE vs. OFAT: Why Changing One Variable at a Time Wastes Runs

Part of our complete guide to DOE for biologists.

If you've ever changed one variable, run the experiment, changed the next variable, run it again — you've done One-Factor-at-a-Time (OFAT) experimentation. It's the default approach in most biology and bioprocess labs, largely because it feels safe: you're only ever changing one thing, so any change in the result must be down to that one thing. The problem is what that safety costs you, and it's more than just extra runs.

The run-count cost is the smaller problem

Say you're testing four variables, each at a low and high setting. OFAT means holding three constant while sweeping the fourth, then moving to the next — a handful of runs per variable, run in sequence. That adds up, but it's not the real issue. A DOE screening design tests the same four variables together, in a single, carefully chosen set of runs, typically fewer than the OFAT sweeps would have taken separately. Fewer runs for the same number of variables is real, but it's the easy part to explain and the least important part of the argument.

The real problem: OFAT can't see interactions

Say your yield only improves when temperature and pH move together — neither alone does much, but the combination does. An OFAT sweep of temperature, holding pH fixed at whatever value you happened to pick, will never find that. It isn't a matter of running more OFAT trials; the design itself structurally can't detect it, because pH never varies while temperature does. You'd conclude "temperature doesn't matter much" and move on, having ruled out exactly the variable that did.

DOE avoids this by changing several variables together in a structured pattern, so the data can separate each variable's individual effect from how pairs of variables interact — both come out of the same set of runs, not two separate experiments.

"But OFAT is simpler to explain and defend"

It is simpler to explain after the fact — "we changed X and Y went up" is an easy sentence. It's not actually simpler to defend, because an OFAT result can't rule out the interaction it never tested for. If a reviewer or a future batch behaves differently because some other variable happened to sit at a different value than your OFAT sweep assumed, you don't have data that explains why. A DOE result, run at a comparable or smaller cost, does.

When OFAT is genuinely fine

If you have exactly one variable you plausibly care about, and no reason to think anything else in your process interacts with it, OFAT and a single DOE screening design converge on close to the same thing — there's nothing to interact with. The case for DOE gets stronger the moment you have two or more candidate variables, which in practice is almost every real optimization or troubleshooting question in a lab.

Start free and run your first screening design — BioDOE picks the design for you from a plain-language goal, no statistics background required to set it up.