There’s no better place to learn about design of experiments (DOE) than your own kitchen. Not being much of a cook or a baker, I do well by restricting my food science to microwave popcorn. Therefore, I happily agreed to help fellow DOE expert Greg Hutto advise his student Jessica Keel how to design an experiment on home-made chocolate chip cookies.
“Want to learn more in your own kitchen? Try making some cookies with different variations in ingredients. It’s a fantastic way to understand and help perfect your signature chocolate chip cookie.”
Danielle Bauer, The (Food) Science of Chocolate Chip Cookies
Optimizing cookies involves a tricky combination of mixture components and process factors. Furthermore, adhering to a gold standard for valid statistical studies—randomization—presents great difficulties. For each run in the combined design, the experimenter must mix one cookie according to the specified recipe and then bake it at the stated time and temperature. It’s much simpler to make a trayful of cookies with varying ingredients and bake them all at once. This can be accommodated by a specialized DOE called a split plot.*
Jessica took on a big challenge: Coming up with not one, but two chocolate chip recipes—soft-and-thick, versus thin-and-crispy. Starting from the specifications for Original Nestle Toll House Chocolate Chip Cookies, she used Design-Expert® software https://www.statease.com/software/design-expert/ to lay out an optimal, combined experiment-design based on a KCV model.** Jessica divided the runs into two blocks to spread it out over her Saturday-Sunday weekend. The experiment determined the effects of four recipe components—butter, granulated sugar, brown sugar, vanilla–baked while varying two hard-to-change process factors—temperature and time—in convenient groups (whole plots).
Jessica cleverly measured the density (by water displacement) and breaking strength (via the ‘penny test’ pictured) of each cookie before handing them over to her panel of tasters for sensory evaluation of taste, appearance and softness on a scale of 1 (worst) to 9 (best).
Focusing on taste alone, this combined mixture-process experiment led to a recipe—heavy on butter, vanilla free—that, when baked at the ideal conditions—325 deg F for 18 minutes—scores near perfect, as can be seen in the ternary contour plot produced by Design-Expert.
See Jessica’s full report for all the details . Then do your own optimal mixture-process experiment to ‘level up’ your homemade chocolate chip cookies. Yum!
*For details, see this tutorial from Stat-Ease that deploys a combined split-plot design to create a “rich and delicious” Lady Baltimore Cake .
**See my webinar on How to Unveil Breakthrough Synergisms Between Mixture and Process Variables.