In the spring of 2022, I did some problem solving for a small but fast-growing, independent cheesecake manufacturer in the Midwest that’s been putting out an excellent line of products for almost 40 years.
But they had an intermittent defect—which is one of the trickiest types of food-production problems to solve. And it had been eluding them for a while.
They’d tried getting help from other consultants, but to no avail…partially because this was during Covid and no one was willing to visit the facility. But this isn’t the kind of problem you can understand without seeing it up close enough to taste, and checking out the production process in person.
So I put on an N95 and took a trip to the cheesecake factory in question. What I found was an impeccably run, gleaming facility filled with proud employees doing their very best to turn out the greatest cheesecakes in the country; I was impressed by everything I saw there. But extreme professionalism wasn’t enough to solve their cheesecake defect.
Based on their own research and analysis, their quality control team had already figured out that the issue was probably sugar bloom—but they didn’t know what was causing the problem nor how to solve it.
What is sugar bloom?
In confections and baked goods, sugar molecules can sometimes crystallize on the surface. Though it isn’t harmful and doesn’t affect taste or mouthfeel, sugar bloom can mar the appearance of an otherwise perfect confection; especially for something like chocolate (cannabis or otherwise) where you want an even, glossy sheen.
Unfortunately, sugar bloom can look like mold to the untrained eye. And that customer perception can be a big problem, leading to returns due to food safety concerns, even though there is no safety issue at all.
In fact, sugar bloom isn’t uncommon in food production, but the cause isn’t always clear. It can happen because of humidity changes, where surface moisture evaporates leaving behind fine crystals of sugar. It can also occur when you freeze a water-based product, because some of that previously liquid water is no longer available to keep sugar in solution.
Before I visited the facility, I explained what my process would entail, including certain costs associated with running a comprehensive experiment (e.g. chemical testing and materials).
They said, “Do whatever you need to do. We have to solve the problem.”
How to solve sugar bloom in cheesecake using science
Luckily for them, you don’t need an in-house professor of food science to apply high-level scientific method to your food-production process.
In fact, I only spent three days on-site in the factory, for fact-finding, experimental design, and staff training on how to gather relevant and impactful data.
Day 1
When I visited the facility and saw the actual cheesecakes with sugar bloom and tasted scrapings from the surface, I was convinced that it was indeed sugar bloom. But recognizing what the problem is doesn’t help you figure out why, and speculation isn’t science.
There are multiple ways to solve sugar bloom, but it really depends upon the exact underlying cause. Just knowing that varying the product’s sugar and water content is relevant is simply not enough.
Even in a recipe with minimal ingredients, it doesn’t make sense to look at just the sugar and water, especially because most ingredients in a cheesecake contain some amount of water, though water itself is not actually directly added as an ingredient. So instead, we looked at each and every ingredient and relevant parts of their production process as potential variables—which is the first step to figuring out how many experimental treatments (i.e. test batches) we might need in order to find and solve the exact problem.
Day 2
We had about a dozen variables we wanted to test. But that could easily get out of hand (in terms of cost and labor) to run the number of experimental treatments required to analyze even that short list of variables.
Fortunately for the cheesecake folks, my PhD work included deep studies of chemometrics—including formal experimental design—uniquely preparing me with an almost comical level of experience, theory, and software knowhow for designing just this kind of experiment. And I did it in a more cost-effective, labor-efficient way than most of my food-science consultant peers would be able to.
Not to bore you with the details, but we optimized the experimental design using a split-plot structure that essentially allows you to efficiently test multiple variables at once, even when some of them are hard to change, and we narrowed our field by focusing on variable ranges that seemed viable and wouldn’t ruin the product flavor or texture.
Then I trained a number of their staff in basic sensory (organoleptic) analysis so they could gather meaningful data on cakes from every treatment.
To make the most efficient use of time and budget, I set them up for testing in a manner similar to consumer analysis (where you ask untrained people to rate a few key sensory factors)—but we used a panel of their team members who were already deeply familiar with the characteristics of the products. Their familiarity allowed us to focus on precision at the level of another method called descriptive analysis, which would normally require much more panelist training (adding significantly more time, money, and other complications).
The result? More and better-quality data gathered very, very quickly.
Day 3
On the third day, the quality control team at the factory started running the test treatments. I was there for oversight, but the goal was for them to run the experiments and sensory evaluation on their own, so they could continue collecting data for weeks after I left the facility.
Using a set of analysis ballots I created for them, they rated various characteristics on a formalized nine-point scale, including liking, sweetness, sourness, etc.
And of course each evaluation must always be:
- Randomized
- Blind
- Duplicated per panelist (at least)
Meanwhile they were also collecting samples of each treatment to send for chemical analysis in the lab.
Going forward
We planned for 8 weeks of data collection, with an initial analysis at the halfway point. Each week they sent me their sugar-bloom data, and in four weeks a clear pattern was starting to appear in the data, so I sent them a preliminary report.
Of course we continued the experiment as planned, and the final weeks of data confirmed the initial analysis. But in less than six weeks total, my client already had an action plan for eliminating the sugar-bloom defect. The rest was just icing on the proverbial cheesecake.
According to the Director of Quality Assurance: “It blew us away, being delivered a root cause backed up by data at the midpoint of the project, along with multiple viable solutions.”
Frankly most food scientists don’t have the background to set up this kind of experiment revolving around sensory evaluation—but the results are well worth it.
More about experimental design in food science
To design and analyze this kind of complex experiment effectively, you have to consider what’s called multiple objective optimization.
In this case we have two main objectives:
- Get rid of the defect
- Keep everything tasting the same for customers.
There may be several ways to get rid of the defect by adjusting certain ingredients—but if they change the flavor or texture of the cheesecake, they’re not very helpful.
To optimize for our needs, we look at the experimental region (using software of course, because it’s multiple dimensions deep at this point due to all the variables) to find the area within that region that solves both our objectives in the most optimal way.
This approach results in much less guesswork after the experiment is analyzed, thereby keeping the cost contained.
Problem solving with food science yields dividends
In food and beverage production, problems and defects will inevitably arise. This is especially true for cannabis edibles where you also have variables like dosage and bitterness. If you want to continue growing, you’ll have to solve those problems. How effectively you do that depends on where you apply your resources.
Whether you’re manufacturing cheesecake or cannabis edibles like chocolate, gummies, chews, baked goods, etc—even minor defects can wind up costing a lot of money in the long run. Especially when you scale up production to meet your growing demand.
Though the problem for my cheesecake client was intermittent, and relatively minor from a quality standpoint—the appearance of sugar bloom could be confused by consumers and/or retailers as signs of mold. Not a good look.
Solving the problem completely and methodically will yield massive returns through the years to come, in terms of minimizing rejected product and keeping customers satisfied—all because of a small investment in forethought and food science.
When it comes to solving food science problems and saving money, more important than any niche knowledge-base is the ability to design an effective, efficient experiment based on the known variables and objectives—so you can run a thorough analysis based upon every last drop of important data.
Once you know exactly why something is happening, coming up with the right solution is…a piece of cake.