Success Story

“Nutritious soda” may sound like an oxymoron. But thanks to some shrewd statistics, PHYZZ is that and more.

A combination of sensory analysis and design of experiments – all done in-house – has helped Cunning Innovations optimize the taste, shelf life, and nutritional value of their innovative new line of sodas. 


ChallengeIn an industry awash with well-resourced multinational players like PepsiCo and Unilever, small beverage start-ups often struggle to get off the ground. Outsourcing R&D to commercial flavor houses is an expensive prospect for small businesses, and the recipe development process is inevitably constrained by modest budgets.
SolutionCunning Innovations, the family-owned company behind PHYZZ-branded sodas, turned to data science to extract maximum value from a limited R&D budget when commercial vendors were unable to deliver a recipe within their target matrix. In-house sensory analysis combining consumer choice and design of experiments (DOE) in JMP Pro, an industry-leading statistical software, enabled the company to optimize the flavor profile, nutritional value, and stability of its beverage line.
ResultsNow available online ( and soon to be in select stores across the country, PHYZZ is a nutritious and affordable alternative to high-sugar sodas. This achievement, says co-founder John Cunningham, was made possible in part through a strategic statistical approach to development that enabled the family to hit their product attribute targets without going over budget. “I can’t even imagine trying to take this approach without JMP Pro,” he adds, noting that DOE and sensory analysis have also streamlined regulatory compliance and laid the groundwork for future product and packaging iterations.

John and DeEtte Cunningham are active people who hike, bike, and relish time spent together with friends and family in the breathtaking natural landscape of California’s Sierra Nevada range. And they are doers – go-getters who see a challenge and set out to find a solution.

One such challenge arrived in 2019 as the Cunninghams began discussing a worrying uptick in public health problems like diabetes, obesity, and poor dental health brought on in part by a high-sugar diet. Unlike most, the pair set out to do something about it. Together with their adult children, the Cunninghams would develop PHYZZ, a naturally low-calorie, low-acid, carbonated beverage fortified with calcium, magnesium and fiber.

What the family jokingly refers to as a line of “un-sodas,” PHYZZ is now available online in grape, orange and lemon-lime flavors, each with only 25 calories per serving. “We wanted to make a product that was truly healthy, with all-natural flavors and sweeteners and low acidity – but that of course also tasted good,” explains co-founder John Cunningham, noting that a key goal was to develop a healthy soda that would appeal to children.

Launching a beverage start-up is no easy task, however, and consumer goods is an industry crowded with well-resourced players. Multinational corporations like Mondelez, Unilever and PepsiCo have immense R&D budgets and decades of experience with food safety, regulatory compliance and marketing.

But no challenge is too big for John and DeEtte. Longtime proponents of industrial data science (John having led analytics transformation in his role as Director of Innovation & Process Improvement at packaging solutions company G3 Enterprises), the two felt that an analytics-focused approach could make up for the comparatively modest R&D budget of their new family start-up, Cunning Innovations.

Designing a recipe to optimize nutrition, taste, stability, shelf life and cost 

In order to launch a line of un-sodas, the first step is to have a recipe. Although DeEtte had developed many recipes over the course of her culinary career, making a consumer product optimized to a set of nutritional standards was another thing altogether. “It’s not easy to make a [mineral-fortified] beverage without any sugar that tastes good,” she says. “Our first big challenge was to get the mix right.”

Desired nutritional attributes included moderate pH, high calcium and magnesium content, low overall calorie count and high fiber. Adjusting ingredients, however, can produce complex interactions affecting flavor, mouth feel and carbonation. Additionally, limiting ingredients to only those that could be naturally sourced created an extra layer of complexity.

At first, the Cunninghams reached out to several commercial flavor houses with their proposed parameters. The samples that came back after an initial round of experiments did not meet their expectations. Further rounds would be required, and the family was not optimistic that the start-up’s limited resources could accommodate an extended commercial development process.

Raw-material cost – particularly the mineral additives necessary to achieve nutritional value – was another potential complication. A recipe reliant on expensive raw materials would invariably drive up the cost to the consumer, and the Cunninghams were on a mission to price PHYZZ affordably.

“We were being told by mineral suppliers that the simpler, lower-cost raw materials would not work with our matrix,” John says. “They warned that the minerals would precipitate or that there would be possibly other stability issues.”

A dynamic tool combines DOE and choice modeling for strategic, cost-effective sensory analysis

As it became clear that outsourcing the formulation process would not be possible, John and DeEtte decided to instead lay out an in-house sensory approach combining consumer choice and design of experiments (DOE) in JMP Pro software. “Mixtures can be really complex, and I can’t imagine trying to take this approach without JMP,” John says. “There are interaction effects and so many other factors we have to take into account. That’s why I just don’t think it would have been possible to do what we’ve done in a spreadsheet.”

With a suite of tools for choice and MaxDiff experiments, factor analysis, segmentation and clustering, multiple correspondence analysis in JMP, and uplift and structural equation modeling in JMP Pro, the software offers not only dynamic visualization tools but a ready-made toolkit for determining product variability and adherence to safety guidelines. Having long used JMP in process improvement, John knew the platform was indispensable across the entire analytics workflow. With JMP, he could design experiments to identify causal factors and achieve an optimal product outcome balancing nutrition, taste, stability, and cost. Design of experiments (DOE), John explains, is the best, most cost-effective way to actively manipulate factors according to pre-specified designs and to really understand the most efficient path forward.

“We used sensory panels to test a variety of metrics including sweetness, sourness, balance, odor and whether the overall flavor was pleasing, mostly using a zero-to-ten ranking system,” says IT and Marketing Consultant Cory Cunningham. Cory explains that consumer data is then modeled in JMP, where insight from choice analysis is used in designing experiments that allow the team to flexibly manipulate components, understand interactions, and identify an optimal design space.

“Statistical solutions have enabled us to extract more insight with a smaller resource bank,” John says. Though their multinational competitors may be able to throw resources at large sensory panels, a smart application of statistics makes it possible to remain competitive, even in the face of very small sample sizes. Furthermore, DOE in JMP enabled the Cunninghams to reduce the number of experiments needed overall.

“DOE is an incredibly powerful approach,” John says. “[As a point of contrast,] the samples that came back to us from the professional [formulation development] companies were nowhere near as good as what our family had come up with. PHYZZ was created not just by luck. It was by statistics.”

Time-based sensory analysis provides insight into stability over time

Another area of innovation for the Cunninghams has been in joining stability analysis with sensory panels to determine not only product shelf life but also how flavor degrades over time. “The product has to last – it has to taste just as good as when it was first canned and have the same nutritional content,” DeEtte explains, noting that stability analysis in JMP Pro has been essential in ensuring that mineral additives were not absorbed by the packaging liner over time.

At first, she says, PHYZZ was bottled in PET plastic with specially designed screw caps that maintained pressure to keep carbonation levels stable. When the company transitioned from PET plastic to cans, they had to iterate recipes to account for packaging effects. “We then used JMP to create a model that would show us how to compensate for anything in the flavor profile that was being affected by contact with the can,” John recalls. “Because of what we found using JMP, we are redesigning how we create our sensory samples for further optimization of current products and the creation of future products.”

Since adjusting beverage recipes to accommodate new, more environmentally friendly packaging, the family has since also built their own filler system that removes oxygen from cans in their production line to help ensure PHYZZ retains its maximum shelf life. “You also have to build a back pressure in the package prior to filling so that you don’t get foaming that leads to variability in fill volumes and carbonation levels,” John adds. “We did some experiments using JMP to figure out what back pressures and purge-gas flow rates were needed to minimize the oxygen concentrations in the package while eliminating foaming and minimizing package fill times.”

Using a statistical approach in JMP will help inspire and iterate future recipes

More than just a development tool, the data generated by statistical approaches contributes to an extensive data history that can be used in future product development efforts. Statistics has been a long-term investment, DeEtte says, that is helping Cunning Innovations remain nimble in a crowded marketplace.

“Way back when we started, we tested so many different combinations that we now have a nice foundation to work from,” John says. “The reason we’re able to come up with new ideas so quickly is because of all the previous DOEs we’ve done.”

The family’s extensive data history has also made the regulatory and compliance process faster and more straightforward, DeEtte adds. “All that data we had is what made the difference [in the regulatory process]. They were impressed!”

“JMP has been so important for this venture. It’s helped us a great deal.”

Try JMP free for 30 days

The results illustrated in this article are specific to the particular situations, business models, data input and computing environments described herein. Each JMP customer’s experience is unique, based on business and technical variables, and all statements must be considered nontypical. Actual savings, results and performance characteristics will vary depending on individual customer configurations and conditions. JMP does not guarantee or represent that every customer will achieve similar results. The only warranties for JMP products and services are those that are set forth in the express warranty statements in the written agreement for such products and services. Nothing herein should be construed as constituting an additional warranty. Customers have shared their successes with JMP as part of an agreed-upon contractual exchange or project success summarization following a successful implementation of JMP software.