Case studies, webinars, and articles you’ll love
Receive the latest content on a regular basis.
Call it baking, smoking, curing, dehydrating, or drying, the vast majority of food processors have to remove water from food. Optimizing your process can deliver significant yield increases—often more than 3%.
In this 30 minute webinar, Scott Campbell shares key insights that will help you achieve significant gains.
Three years ago, on a foggy October morning, I made a visit to a food processor that I will never forget. I had gone there to do customer research.
The staff members and I suited up and picked our way through dark corridors until we found the quality lab, where our AQUALAB water activity meter sat alongside a bank of moisture analyzers. The QA manager then pointed at our water activity meter.
“That’s the only reading I care about,” she said. Then she pointed to the bench that held four moisture analyzers.
“I run these machines because one of our owners demands that printouts of the moisture content for each batch be sent to him at the end of every business day.”
Dear listener, because METER is a water activity company, I wasn’t pleased to hear that. What I’ve realized since then is that this owner is actually a genius because he knew that the moisture of each batch drives profitability for his whole business.
That is the first point we will discuss today: Most food processors don’t understand what mistakes in the drying process actually cost them. This guy did understand that, but that’s not why I remember him. I remember him because even though the QA staff had been sending him this data for the past 15 years, their average moisture content had not improved over that time.
It dawned on me that if I could figure out a way to get QA managers, plant managers, continuous improvement managers, technical services directors, and business owners, all on the same page, then every food processor could become a drying superstar.
That is the second point we will discuss: How to fix the mistakes that prevent you from becoming a drying superstar.
Let’s begin by answering a simple question. What is drying? By this we refer to any process that takes water out of a product. This could be baking, smoking, freeze drying, curing, spray-drying, cooking, evaporating, all are forms of drying. And by drying, we include both batch and continuous processes.
How does drying variability impact the performance of food processors? Variability is the enemy. It gives you over and under dried batches. These batches result in lower yield, lower throughput, higher energy costs, moldy products and dissatisfied customers.
In spite of this, I’ve learned from visiting hundreds of food processors that every one of them has a lower average moisture content than they need to, because they’re trying to avoid the catastrophic costs of underdrying their products.
The relationship between drying consistency and average moisture content is easy to see by reviewing this graph, which shows two histograms, one with a high standard deviation, and one with a low one.
If your absolute threshold for batch failure is 14% moisture, then it’s easy to see that the company with the lower standard deviation will be able to produce at a higher average moisture content.
Here’s a map that plots a journey from wherever you are now to becoming a drying superstar.
It shows how reducing variability in your drying process is possible over time. But why embark on this journey? Why try to reduce the standard deviation of your drying process? Because failing to do so is literally costing food processors billions of dollars each year in lost value.
The cost of drying variability for any customer is easy to calculate. I just need to know a few things about your business.
First, if you made one more unit of product, could you sell it for a market price or would it be worthless? This is important because if you can sell that unit you make, any increase in yield is multiplied by the market price of what you’ve produced.
A good example here would be almonds. If you make circus peanuts, however, the world can only eat so many of them, increased yield only reduces your cost of goods sold.
Second, I need to know how much your average moisture content could increase if you did a better job of drying your batches.
Third, I need to know how much product you make every year.
Would you believe if I said that for a relatively small food processor, drying process success can be worth between $6 million and $9 million?
Let’s compare two companies that make circus peanuts.
Patches runs a consistent process, but Bozo does not. For the sake of comparison, we will say that Patches’ company and Bozo’s company both have identical revenues of $30 million a year. We will also assume that if circus peanuts are made at a moisture content of above 14%, they will mold in the bag.
Patches can run his process at an average moisture content of 12%, while Bozo can only run at 10% because with the standard deviation of 2% moisture, he must run at a lower average to make sure he does not have any under dried batches.
This is a 2% difference in yield. How does that translate into profit?
The average cost of goods sold for food processors in the United States is 67% of revenue. Because their yields are different, Bozo must spend $20 million to fill up his orders, while Patches spends only $19.6 million to make the exact same amount of circus peanuts.
An average food processor has operating margins in the United States, again, a 5.5%, which makes Bozo’s profits $1.67 million. Patches’ profits are 24% higher at $2.07 million.
Most publicly traded food companies’ stock price and company value is about 15 times their operating profit. That makes Bozo’s business worth $25 million. Not bad, but then consider that Patches’ business is worth $31 million — a difference of $6 million.
This calculation does not factor in the cost of quality losses, extra energy used to drive off moisture, machine time used up by batches that were in process that should have already been finished or extra labor time used to manage those batches. That’s just what was lost by drying off water that could have ended up in packages.
If we do the same calculation for almond producers, the difference is even greater.
This is because our two businesses, which we’ll call Nutjob and Nutsanity, can sell any increase in yield at market prices. That makes Nutjob’s revenue $30 million, but on the same inputs, Nutsanity’s revenue is $30.6.
Their average moisture contents are different, just as they were in our prior example, but their annual production is different because Nutsanity can sell any extra almonds that they produce.
They produce 120,000 pounds of almonds that Nutjob could not. Almonds sell for about $5 a pound. Thus, all the extra almonds that Nutsanity produces, raise their profits to $2.27 million, which compares favorably with Nutjob’s annual operating profit of $1.67 million.
When we now consider what Nutsanity is worth and what Nutjob is worth, Nutsanity is worth a whopping $9 million more.
Because drying variability destroys so much value, we need to know what causes it.
Food processors make two huge mistakes when it comes to drying: They use the wrong drying metric, and they don’t collect process data and write it to the batch record.
Let’s first talk about using the wrong metric.
By far the most common method for measuring moisture content at food processors is rapid loss on drying moisture balances.
These are the data that end up collected in the batch record. These moisture balances look like this.
Moisture balances dry products and measure how much their weight changes when they do it. As far as lab instruments go, they’re relatively cheap, costing between $1500 and $5,000.
This is where the good news ends. Rapid loss on drying moisture balances are slow, inaccurate and perpetuate bad science. They take 10 minutes or more to read, have a precision of plus or minus 1% moisture. This means that the standard deviation in measured moisture in an industrial process is sometimes mostly due to the measurement method alone.
I could do a whole virtual seminar on why moisture balances don’t work well for measuring drying process success, but we don’t have time for that today.
Instead, I’ll just say that there are three types of listeners on today’s virtual seminar: those that don’t measure anything about their drying process, those that use water activity, and those that use moisture balances. If you’re part of this last group and are convinced that your measurement method is perfect, stay tuned because I have a graph I want to show you later in this presentation.
The second mistake that food processors make is that they don’t have good systems to record process data and write it to the batch record. Without such a system, the effect of process variables on moisture remains a mystery.
There are three types of process data you need in your batch record: Ingredients, environmental data and machine settings.
We’ll start with ingredients. Do you get a certificate of analysis with moisture content on it from each of your ingredients suppliers? Remember that even if you do, it’s unlikely that this is a useful process variable because of the shortcomings of moisture balances we just talked about. Also, note that if you don’t know the moisture content of what suppliers are sending you, your process could be sabotaged by ingredients that are too wet or too dry.
Incoming ingredients can also refer to pre-drying process steps that may happen in your factory. For example, we have found a clear correlation between meat slice thickness and drying times for beef jerky. Failure to collect that data for each batch is a missed opportunity for drying process improvement.
The second factor that affects your process is your factory’s environment. This includes temperature and vapor pressure, as well as post processing storage steps in which moisture may be gained or lost. Remember that your yield depends on what makes it into the package, not what your moisture was when your product emerged from the dryer.
The third source of data which should be written to your batch record is machine settings. The changes that an operator makes to these machine settings are critical to the drying process. Take a moment and think of the best machine operator in your plant. How does he or she compare with the other operators and staff? How much better is that one individual than the average?
Chances are that person has earned a sterling reputation by successfully running a complex process. He or she changes machine settings whenever necessary to keep the process humming along without a hitch.
What isn’t clear, sometimes even to your superstar operator, is how he or she does it. Parameters like dwell time, steam settings, belt speed, and oven temperature can all be changed, but their correlation to moisture often remains a mystery.
What’s more, even skilled operators may not have the right incentives in place to achieve success. For example, we have visited pet food factories where operators are running at moisture levels that are 50% too low because they’ve been criticized in the past for causing mold issues.
Fixing these mistakes can seem like a daunting task. We at METER Group have spent the last five years building tools for nailing the moisture up every batch you make. Here are our recommendations for making an immediate improvement in your drying methods.
If you do nothing else based on this virtual seminar, get a better yardstick by replacing your moisture readings in your process with water activity.
Moisture content is a measure of how much moisture is in a sample. Water activity measures how much energy that water has.
We understand that in some cases this is more difficult to understand than moisture content. In case you’re not familiar with water activity, we measure it by placing a product in a closed chamber and measuring its equilibrium relative humidity. We can talk about the science of water activity and how it compares to moisture content, but that’s a different virtual seminar. Today we only have time to talk about why it’s better than moisture content.
There are three reasons why this should be the parameter that is recorded in your batch record.
Here’s that graph that I promised you water activity skeptics.
It shows the relationship for a dry product between water activity on one axis and moisture content on the other. Moisture content here is on the Y axis. We can see that as the moisture content increases, so does the water activity.
For dry product and in the range that most intermediate and low moisture products are, this curve is relatively flat. If we zoom into the range that I was talking about in our earlier example from 10-12% moisture, and this is the uncertainty of a moisture balance, we see that the precision of a water activity meter at plus or minus 0.003 water activity units is an order of magnitude greater than a moisture analyzer. There is simply no comparison between the precision of the two methods.
One more point about keeping a complete batch record of product water activities. Dear listener, it is the 21st century. We live in the age of iPhones and Tesla autopilot. I get incredibly frustrated when I visit food processors and see they’re still using pen and paper, forever trapping data in three ring binders that will never help them improve the success of their drying process.
I am so frustrated by this that I’m announcing a free version of SKALA that is available to everyone on today’s virtual seminar. All you need to take the first step on your journey to drying Oracle status is an iPad, an AQUALAB, and a small box we will send you for free if you use the code NOVICE20, 20 being the year you started your journey to becoming a drying superstar.
Here’s how it works. Once you have the system set up, on the iPad you click on the name of the product you’re testing. Then every time you take a water activity, it’s written permanently to the batch record.
One user was able to track their continuous oven in near real-time and significantly reduced variability just by making this data visible to operators in her factory.
Here is an actual graph from that customer that resulted when she started tracking the batch water activity every day.
Here’s a view of that same water activity plot for that product three months later.
This user was able to reduce moisture content and water activity variability by 30% just by showing the data to her operators. For clients that want access to the next level of drying expertise, METER recommends characterizing your entire production process and writing variables to that same database that contains your water activity information. METER offers a paid version of SKALA that does this.
This includes incoming ingredient moistures, environmental parameters and machine settings. An embedded statistical process control tool then enables clients to make a list of variables ranked in order of their correlation to batch moisture. This is the right place to start any process improvement.
Keep in mind that once you’re writing this data to your batch record, this becomes a permanent asset to your company. But also note that your process is complex and unique. It can be characterized, but it will take effort and commitment on your part to continuous improvement. METER has a professional services team that can make this process easier and your reward for doing this will be insights that you couldn’t have gotten any other way.
Now there’s just one more thing that I want to tell you, and that is once you have achieved expert level status as a drying superstar, your attainment of Oracle level status is already guaranteed. And that’s because within SKALA, we have embedded machine learning algorithms that improve the success of your drying process day in and day out.
How do we do this? We start with the physics of drying. Do you need to know or understand any of these equations? No. They are all built into the machine learning system inside SKALA.
It is important to know however that our machine learning approach is not a black box. Here’s how it works.
Using first principles and under conditions, measuring inputs like the production environment and your incoming ingredient moistures, what machine settings will yield the best water activity on each batch? That is what the machine learning algorithm asks each day.
For example, a manufacturer of meat snacks that are smoked in a batch oven recently implemented our SKALA system on its manufacturing floor. The system started with a model based on the physics of drying food, and then used meat thickness, oven wet bulb, oven dry bulb, and meat internal temperature as model inputs.
Each time the drying process was completed, the machine learning model in SKALA had a predicted water activity and moisture content for that batch. Technicians then measured batch samples for water activity and compared predicted to actual moisture levels.
Each time this happened, not only could the clients see how well the model was working, but the model learned gradually improving it’s R squared and root mean squared error of predicted moisture values.
These data can be seen in the graphs on the left of this slide. The information depicted here shows how much that model improved in reduced error and a higher predictive capability over the first 60 batches that the customer ran. Note that each time a batch is run, that new batch improves the predictive capability of the model and now the system alerts operators when drying processes should be stopped to achieve target moistures. This has reduced the standard deviation of batch moisture content by 55% after three months.
Please don’t get hung up on the fact that this customer’s drying process control variable is dwell time. The system works whether your control parameter is dwell time, belt speed, oven temperature, steam settings, or anything else.
So let’s finish where we began.
That customer that I mentioned at the beginning of this seminar that I visited three years ago, who required moisture readings each day, today they’re a SKALA customer.
They use water activity in their batch record. They record the parameters that lead to that water activity and I’m happy to report that they’ve reached an expert level of drying in only one year. They still have a ways to go, but getting this far is something that we, that all of us at METER are deeply proud of.
Here’s a quick summary of what we’ve discussed.
You mentioned there was a customer who improved 33% while using the free version of SKALA. Were they just tracking water activity? What were they doing there?
Yeah, so that customer was using a paid version of SKALA because what we had done was set up all of their water activity meters and other process parameters to be on a live dashboard that was displayed to operators. So they were using a paid version because our pro services team had to go in there and install that whole system and support it.
They could have used a free version and gotten some of the results, but in some cases, having that data displayed and having the system supported long-term is critical to achieving those results.
So I believe that many listeners on today’s virtual seminar can achieve improved results just with the free version, but that particular customer had paid our pro services team to come in and do a complete install.
What’s all this cost, Scott?
Great. And that’s a great question and we do get it a lot. So the cost depends on how deep you want to go with the install, how many data sources you’re integrating into the batch record.
A very simple rule of thumb. And in terms of costs, we understand that food processors are very price sensitive. We talked earlier on in the seminar about the average operating margins of food processors being 5.5%. That’s slim. So a good way to think about that is by comparing to other approaches for reducing variability, which might be buying new production lines, better machines, smart machines, things like that, those will cost millions of dollars. The best metric to think about how much this will cost you is look at that at whatever you’re losing in not being a drying superstar.
So what is that number? And for nearly all of our customers, what we’re charging them is about 10% of that value. So looking at an example of those almond producers, one of them has an opportunity to gain $600,000 a year for a system, a full install that integrates all of their process parameters, all of their water activity data, all of their factory production data, all of their incoming ingredients, everything, that will cost one-tenth of what they stand to gain in terms of increased profit.
And I often get the question, “Are you willing to guarantee that?” Yeah, we are. We’re willing to trial it with you and we’re willing to guarantee it. We know that not achieving an increased profit for you, doesn’t do anything. So a good rule of thumb is one-tenth of what you stand to gain from implementing a better drying system and those are things that we guarantee to clients.
For a process like drying cannabis, where you don’t have access to continuous monitoring of environmental controls, how effective would SKALA be?
So what we’ve experienced in the cannabis industry is that we do have access to those data. So let’s say that your drying rooms don’t have control in them right now. At the very least we would put a temperature and RH sensor in those rooms and be tracking that over time. Just those two data points are enough to give us a live prediction of what the moisture content, water activity of each batch in that drying room is. So the system doesn’t work without process parameters, but even in cases where your drying rooms don’t have them, they are very easy to access.
Does SKALA interface with any food safety programs software for compliance with FSMA and HARPC plans?
Yeah. So there are two parts to the answer to that question.
The first is that yes, SKALA does have an API that allows for data to be sent and digested by other systems that could be an ERP, it could be a food safety program, it could be a manufacturing execution system. SKALA can interface with these systems.
But one important point is that SKALA itself and we haven’t talked about this at all today, is a 21 CFR Part 11 compliant system, meaning that if you are making a product and you need to log that the kill step was reached, we’re capturing those data from your production process. We’re writing them permanently in a 21 CFR Part 11 compliant software system and they’re available for internal auditors, third-party auditors and food safety managers to verify each day before batches are actually shipped. So SKALA itself does function as a food safety system. It’s just not what we were talking about today in our seminar on drying processes.
Case studies, webinars, and articles you’ll love
Receive the latest content on a regular basis.