A tale of 4 friends…

Four friends were discussing over a beer how to build a model rocket that would soar higher than the rest. They had quite different ideas about how to approach such a challenge and each one of them thought theirs is the best. They entered a bet and wanted to meet again in three days to find out who was right. Each of them prepared with the best rocket they could build.

Randy Random loves to try whatever pops into his head. More engines? Sure! Different fuel? Why not!

Sally Sequential is for more structured approach by changing one thing at a time. First the fins, find the right shape, then the engine, then the size...

Danny Detail is afraid that when you don’t test every possible combination that you won’t be able to find the best parameters for the rocket.

And then there is Evelyn Efficient: She just took a DoE course and knows that neither one of them has a chance of winning.

On launch day, Randy's rocket exploded. Why? He doesn’t know. It worked fine when he tried yesterday.

Sally's shows a decent flight with a decent height. She clearly found an optimum in terms of parameters.

Danny is still testing. No one of the friends has seen him in the past three days.

Evelyn's rocket shoots up, higher and higher until they lost sight. Higher than the rest with a beautiful balance of power and stability.

Randy’s intuitive approach

Randy‘s approach is unpredictable. It might have worked well yesterday but he was probably just lucky with the wind.

Sally’s One Factor at a Time approach (OFAT)

Sally‘s approach is already quite good and she also found some decent parameters. It’s probably the way most people approach experimental design today because it is intuitive. If you find the optimum for each parameter individually, you find the overall optimum right? Wrong! Sally found a good solution. But she missed the best solution. Because the factors in an experimental design are often not independent. They interact with each other. Consider the fins, they might need a different design depending on how powerful the engines are.

Danny’s detailed apprach

Danny, well… He would have found these interactions but… it‘s just taking forever.

Evelyn’s factorial design approach to DoE

That’s where Evelyn‘s DoE approach comes in. In a nutshell, DoE aims at the most efficient way to structure your experiment to get the most out of it. There are a lot of different preset design plans you can choose from to match the specific problem you want to solve. For example, Evelyn uses factorial designs to test multiple factors at once. She's not just changing the size of the rocket; she's also tweaking the fin shape and engine in a structured way. By doing this, she can see not just what works best, but how different factors interact with each other – something the others miss. And all that with much less experiments than Danny had to do.

Here is a list of design plans that you can use for different things:

• Full factorial design: Testing every possible combination of the factors studied. Requires a high number of experiments, which makes it expensive.
• Fractional design:Reduced number of experiments containing only the most important features.
• Plackett-Burman design: Screening design to identify the most important factors.
• Central composite design (CCD): Good choice for optimization problems or finding robust process parameters.
• Box Behnken design: Similar to central composite designs but slightly different structure.
• Latin Hypercube design / Latin hypercube sampling: Interesting methodology to explore certain design spaces, especially for high throughput screening (HTS).