From Excel Hell to Automated Production | Ep 7

Mikkel Svold (00:00):
How do you take the journey from low integrated to fully digitalized?

Linnet (00:06):
Do you need to establish what is your ground zero so to speak. So where are you coming from? What do you have in place today? And even that can actually be quite a challenging.

Mikkel Svold (00:16):
Are they already matured or no?

Søndergaard (00:18):
A lot of companies need flexibility and need to do something fast and then they start manual.

Mikkel Svold (00:25):
So would you be walking around the plant and kind of saying, “Okay, we need this and that connected.”

Søndergaard (00:30):
We need to understand how everything works.

Linnet (00:34):
A large part of this is also understanding their workflow. We go around with a lot of different panes. We can also identify and ask them questions.

Mikkel Svold (00:40):
What numbers would a production manager want to monitor?

Linnet (00:44):
Very simple thing, just like are the labels you’re printing actually the correct labels? But we are producing so much volume so quickly, printing the wrong label and putting it on, they can be hours of work of correctifying and reworking that.

Mikkel Svold (00:59):
Hello and welcome to the FoodTech Podcast. I’m Mikkel Svold. And today we are going to talk about how you can actually take the journey from a low integrated food and beverage production facility to something that is maybe even fully digitalized, fully automized. How can you take that transition and how can you make data available for the right people at the right time? And how would you collect it? So it’s a big thing that we are talking about today, but I think it’s going to be very exciting because with me, I have Erik Søndergaard who is… I was just about to say the financial brain in this little setting. So you represent basically the needs and wants from management. And I also have with me again, Lars Linnet, welcome to you also. And you’ll be then the engineering brain, so that be the one who knows about what shop floor people would want, what does the operator need, what is necessary for controlling shop floor?

(01:57):
So I think it’s going to be a very exciting conversation, because I’m sure that the data that you can collect is one thing, but the results you want to see on the screen and the numbers that you are actually looking at and measuring, those are very different from each other. When you talk operational people and when you talk management people. You’re both nodding, so I hope that’s more or less right. I think actually just diving right into it, so how do you take the journey from low integrated to fully digitalized? What does that journey entail?

Linnet (02:41):
Well, I can start off saying that first thing of course you need to establish what is your ground zero so to speak. So where are you coming from? What do you have in place today? And even that can actually be quite a challenge actually, to really understand sometimes also for the customer actually what is their own setup exactly and what dependencies do we have and what’s the technical state. And then of course you need to have a strategy, a vision. Where are you going? And sometimes that can be very concrete. That can be we increase capacity. We bring in a new line. And other times it can be just we need to be compliance with a cybersecurity issue or something like that. Or generally we just need to improve the performance of a plant generally. So I think that you need to have that a cost and an effect so to speak.

Mikkel Svold (03:39):
How often do you guys visit a beverage production plant? It could be a dairy, it could be anything else. It could be juices, something like that. How often do you visit a place where they want to do this transition? Because to me it seems like it should be a transition that should have taken place already. Are they already matured, or no?

Søndergaard (04:02):
If you look on a typical mid-range company, a lot of companies need flexibility, need to do something fast. And then they start manual and then a lot of Excel spreadsheets arrive because you want to have more numbers, perhaps more compliance. And suddenly the sales department sells a lot. So you have soulful capacity and what do we want? Do we want to invest in a new building? That’s a big step. Or could you optimize what you have? Perhaps labor costs instead of moving to China, then you keep local, but then you need to optimize more about the production, so one person can run four lines or something like that.

Mikkel Svold (04:52):
I think before we turned on the microphones, you told me that the two of you, you just recently visited a production plant. Can you try and take me through that visit? What happens?

Linnet (05:05):
Well, we went there. This is a medium-sized food producer in Denmark. And just like Erik just outlined, they went from this producing something, having success, growing, and being at I would say a relatively okay state with doing manual processes and Excel spreadsheets and managing like that. But all of a sudden it becomes the glass ceiling for them for going into the next growth with what they have. Because this company was also geographically placed where it wasn’t so easy to recruit people, et cetera. So they had a lot of challenges actually. So we went there, talked with them, interviewed the people. They had some wishes of course on their own. And when we used our expertise to say like, “Okay, but maybe you have these three low-hanging fruits, maybe you should also consider taking these in on the project.”

Mikkel Svold (06:09):
Would you be walking around the plant and kind of saying, “Okay, we need this and that connected and we need data collection point here.” And how do you identify those opportunities?

Søndergaard (06:24):
We need to understand how everything works. And as we also talked about when you are going around, you hear a lot of things, how people talk to each other, which questions do they have? And when you go to the different monitors, you see a lot of posters-

Mikkel Svold (06:46):
In the office buildings.

Søndergaard (06:47):
Yes. But you see posters on the screen or on the operator. And why are the posters there? Is there something they should remember? There’s a lot of things to talk about to get an understanding where they are.

Linnet (07:01):
Yeah, because I also think a large part of this is also understanding their workflow. And then of course we go around visiting a lot of different sites and see a lot of different pains and so we can also identify and ask some questions, say like, “I know the cabinet, that product is end of line. How do you actually manage that?” It is very much going around using your eyes, using your ears.

Mikkel Svold (07:30):
And now in my introduction I said that when you talk to management and when you talk to operations, they will require different kinds of numbers. They have different KPIs. They’re looking after different things. How does that become evident and what sorts of numbers are we talking? If we start with management, if they’re not already, what should they be looking at looking for?

Søndergaard (07:56):
In the management we have for example, very important customers and we say we will deliver 99% for example. Can we do that? How is the production going today? Because typical what you produce today needs to be sent at the end of the day. Is the production going 100%? And every day something happens. So to have a dashboard, how is that compared with the order for example, to compare these things? It could also be that we invented a new product and we need to buy in raw material, so we need to buy them before they make production. But when they’re running the production when we need to buy in more because we don’t want to buy too early. And now if you don’t have this integration you need to count every week and use manpower to see how much is in stock, because yeah, it’s manual.

Mikkel Svold (08:55):
Yeah. So management, they would basically look at how are we performing held up against what we promised our customers and delivering what we said.

Søndergaard (09:05):
Yes and what’s the cost of the product and so on. So we need in a higher level, but we need the numbers to be the right ones.

Mikkel Svold (09:15):
And then I look at you Lars, because where does those numbers come from?

Linnet (09:21):
Well I think for a good example could be, let’s say you have operator standing out on the line producing a job order that he has been told from the office that this is what we need to get out the door today. And when something happens on the machine, maybe a breakdown or a delay or something like that. Then what’s important for him is that the guys in the office of the management, they can adjust and take corrective action. So either we move that production order if it’s higher priority to another production line, or ultimately of course delay the order. So I think what it could look for him as an operator… Oh sorry, maybe I lost track of your question.

Mikkel Svold (10:04):
No, it’s just those data that you deliver to management, they come from production.

Linnet (10:09):
Yeah. And those will then be tied in with the job order. So typically the machine or the production line should then deliver the job order, the number, what we’re producing and the target number and what we currently have produced. And then if we are running… And then sometimes-

Mikkel Svold (10:29):
So basically if production is on time.

Linnet (10:31):
Yeah, exactly on track and sometimes you can have a projected completion time et cetera. You can get more advanced of course. But yeah, that would be the data that the shelf floor event deliver to that system.

Mikkel Svold (10:43):
And then one thing is what management needs, what numbers they want to monitor. What numbers would a production manager want to monitor? What is necessary for the operations?

Linnet (10:58):
I think that when we start up production, first of all it would be for example, I think it would be a great help that if you are… Just very simple thing, just like are the labels you’re printing actually the correct labels. That may sound actually, but we are producing so much volume so quickly a lot of these places that even if you are printing the wrong label and putting it on, that can be hours of work of correctifying and reworking that. So I think that for the shelf floor for example, that could be a nice integration point to have a camera that validates the label that has been printed and everything is okay. And of course you can’t do that without the integration and how to put that in place.

Mikkel Svold (11:41):
I’m saying that camera needs to know what to look for. I know there’s not the camera doing the brain work on that, but somehow it needs to know what batch are we talking right now and what are we trying to produce?

Linnet (11:52):
Yeah, exactly. Because it would be pulling the data from some centralized place, for example MES system. And then making a compare and then telling the machine that everything is okay, we can run. Or more importantly, if there is an error on the label, then stop it and make sure that the operator gets notified about it and it can be correctified.

Mikkel Svold (12:13):
And then when we are talking a not highly digitalized production plant with a plant that has low integration between these different layers in the production cake, so to say. What do you do? What’s the actual, you mentioned an Excel spreadsheet, is that how you control it.

Linnet (12:43):
You mean that would be your base point?

Mikkel Svold (12:45):
I mean how do you make sure because these validations you still need to do, even though you don’t have the high-speed camera photographing all the labels.

Linnet (12:54):
It would be the packaging people maybe who are managerially handling it, at one point they will probably notice before it gets sent out the door.

Søndergaard (13:04):
They get a production order, we need to make this production. And they have several production orders for one day and they sort them perhaps how could we make the production the best way? And then they start production, put in, “We need to make this product with a recipe.” And then they type in how much amount of the different material they need to use for that production.

Mikkel Svold (13:30):
How much flour do we need for these rolls? How much water does it need to be added? And they would type that in manually?

Søndergaard (13:37):
Yes. So they have the recipe and they, “Okay, this amount,” because if it’s a big order or a small order, they will put in those numbers. And sometimes during the year if some of the raw materials is perhaps tomatoes or oranges, they have a different how much water based in it. So during a production year you need to make modifications of that recipe. So do they remember to do that to look on the paper is could they remember for yesterday? So when you buy something in the market supermarket it should taste the same but the raw material are changing over time. And that’s the difficult part of it. So from the management perspective, we could have a customer say, “Okay, we want another product where we want to have a little bit more of apple in for example.” And then you need to change the recipe. So if it’s automated, we have made that, “Okay, you can do it at the order sites and then go automatically to do it the right way.” If it’s not, then I will look at Lars and you have a production and you have a man.

Linnet (14:51):
Yeah. And then you have a risk of someone just doing human error and typing in the wrong value.

Mikkel Svold (14:57):
How often does that happen?

Linnet (14:58):
I don’t know how often. But it can happen so it does happen, just because the quantities are so high. And if you are unlucky, you actually ship that out the door and to the customer, so it can be end up at the retailer and then they have to send it back or something like that. So that is really one of the things also we are looking for when we come out visiting to see do we have a recipe system in place? That is because in my view there’s so much value to that functionality to get that implemented and it’s something that we know very well. And usually the recipe is already well-defined. So getting that in place, of course that can be an easy task, it can also be a bit more challenging task to do it. But nonetheless it is absolutely doable.

Mikkel Svold (15:53):
Let’s dive into that actually. So what kinds of data would be interesting to log in a system? And the system we’re talking about here is the MES system. Not all companies they have one, because they can handle it also via Excel spreadsheets like you mentioned. But what sorts of data and information would be clever to log in that kind of MES system? So you mentioned recipes, what else?

Linnet (16:22):
Well, I think especially for the end-of-line equipment like the packaging machines, there’s a whole standard called PackML that defines the data exchange for the MES system. And what you exchange there is the batch ID, like what your water you’re running the produce quantities, maybe even machine speed, the stoppages courses. So you can for OEE numbers optimization, it can be critical process values like temperatures and flows, et cetera. Like values that tell you something about the production conditions that are critical and they may impact the final product quality.

Mikkel Svold (17:12):
And would there be any managerial information or data that needs to go into such a system?

Søndergaard (17:18):
Yes, the material you used, what’s best for where you have bought it. Because sometimes here in the radio that they call a product back because there was something that was not good. So you need that traceability as well. So if you have an automated system, we can see what has been put in and where did it move from point A to B so we can follow it through the line instead of making it on paper so that you have very short time to when you need to document that you can make this traceability. And when it’s done automatically, yes then it’s done the correct way and the operator can do something else and be more effective on that case.

Linnet (18:13):
And I also would add to that that you also have a warehouse management when you incorporate that into this. So let’s say you are having a filling machine that is consuming cups for the product, then you can in real time update and reorder cups for let’s say packaging machine. And the operator can even get notified that now he should go out and get a new pallet of cups because he will run out in 30 minutes.

Mikkel Svold (18:43):
So I want to take this back to the plant manager and to the operations manager. So what would be the low hanging fruit? You mentioned the recipe system. What would be the low hanging fruits to implement? I’m saying tomorrow, but I mean once you get the system up and running. And also what does it require to go from Excel hell to automated system? What components do you actually need? Because there’s also, I’m guessing a hardware bit and you need software that can run it. So how do you get from zero to something? And what is that something?

Linnet (19:32):
I think we should start at the shop floor. You can do several levels of integration. So let’s say in the PLC and the SCADA, which is the main screen from where the operator is controlling the plant. Maybe a significant step would just be that he gets a job order with production list in the list there and he can see. And then he can preselect and say, “This is what on running.” And when the machine starts up, it requests the actual recipe. And then it gets downloaded and we make sure that the recipe that’s been downloaded into a machine is also valid and nothing went wrong. And then we also get the back flush that now we have produced XYZ amount of it. It can also be more automated. So actually basically the system would just be running an auto and just eating away of a job order. So it has completed job order A, and then automatically it just transitions to the next job order without the operator having any hands on.

Mikkel Svold (20:36):
So in my head what happens now is that I picture me sending a PDF file to my printer and then another person here at the office sending another PDF file to the printer and then that printer will just automatically go through that in that order.

Linnet (20:50):
Basically that is what will be happening, yeah.

Mikkel Svold (20:52):
If you print food.

Linnet (20:53):
Yes, exactly.

Søndergaard (20:55):
3D Printers.

Mikkel Svold (20:57):
3D food printers.

Linnet (20:58):
And you can say what would be needed from the structure from a control system would be that the PLC would need to have some kind of connection to the MES system. And we would also need to agree upon a format on which the protocol, how should this data be structured and what do we mean when are we downloading the recipe, when are we starting, when is the job order complete? And all that is actually detailed in the ISA-88 and S-95 standards.

Mikkel Svold (21:30):
So those were the standards that we talked about in the previous episode on standards, exactly. Okay. I’m thinking how do you actually take these say that we do these automizations and we get these different KPIs, we get numbers. How do you present the numbers so they are understandable for people?

Søndergaard (21:59):
Yes, you can do it in different ways. And again about money and how much time do we want to fill in. Because if you said, “Okay, I only want to be compliant, I need a report that I have made this product and the temperature has been within the needed numbers.” Then you perhaps print a report, put it in… Yeah, wherever

Mikkel Svold (22:22):
You put it on the shelf.

Søndergaard (22:23):
Yes, and then it’s done. But what we see is that there is money in data. Because the data shows you how can you improve, how can you make statistics? And you have done that several times from CFO. We want to do it in the production area. And what you see is when things happen. So we make a lot of dashboards. We say, “Okay, we have some KPIs. So if they every morning talk what happened good yesterday and what went wrong on a packaging line for example. Where is top ten?” And the data is ready and they could talk about, “Okay, the one stop that took five seconds but we have 30 of them could actually have a big impact.”

(23:06):
So instead of they’re missing because you didn’t check the numbers in Excel, you get facts. So to have a stand up meeting where people can look at that numbers and no effects there, no arguing, it’s, “How do we fix it?” You will continuously improve that packaging line. So that could be one area. It could be another area and say, “Okay, can we optimize the production and make them in another order to have less cleaning time?” But to have it on the big picture, because if it’s one person doing one job and the next who decides what to run when and is there two orders almost the same product to two different customers that could be collected together in some part of the production. To have that overview is where the money is to optimize.

Linnet (24:07):
But I want to say it is also a challenge always to present something that is complex in a manner that doesn’t generate information overload at the operator. So therefore like said, doing dashboards and do something that it’s graphically nice. It really has the most important information on top is really key to have that. Because you will see a lot of applications where people didn’t adhere to that. And it can be quite a challenge sometimes actually to read and fully understand the value of a data presented.

Mikkel Svold (24:41):
Yeah. I’m also guessing that when you have an operation that has been running for years and years, without all the data you must have operators. Even people in management, you’ll have people at every step at every layer of the organization that has a little resistance to this. How do you work with that? How do you make people trust what they see and act on it?

Søndergaard (25:15):
Yes, it’s like every change. And one of the most important thing is to evolve. And they have typically–or I would say always–a lot of knowledge that is so important to take in. So if you just come from the top and say, “Now it’s this way,” you miss so much. And it’s also when we are looking on the numbers that if you should use it, it’s correct numbers. I’ve seen several times that you want to make a salesperson come in, show some dashboards, “Oh, they look really nice and you buy in.” But the data is not available, or is it the correct data? So then you get garbage. So you need to have the operator as a part of it. And you need to understand the language to make the collection in the correct way and then they can be useful. And that’s why we are too sitting here because you really need that.

Linnet (26:19):
And I also think address the pain, obviously. Evaluate the need and really highlight what adds benefit and in what way can the solution help them today. It is really about trust and really coming in. And I’ll actually say sometimes it’s actually easier to come into a customer which doesn’t have anything and then you can present solution, compared to come into someone who has maybe a legacy system that they’re holding onto. And then sometimes they can be for different reasons, a lot of pushback because they really like the old system, they trust it, they know how to operate it, they know how it looks. So it is also, it can at least be also a change management project.

Søndergaard (27:05):
Yes. But try to imagine that you are an operator and you are put in the recipe and you then press the button and know it’s €100,000. “If I made a mistake, it’s €100,000 or €10,000.” Whatever. You need to be really sure.

Mikkel Svold (27:27):
It’s a big responsibility.

Søndergaard (27:28):
Yes. And you make shift new people coming in. And then you certainly see a system where, “Okay, what the customer has ordered is automatically there.” You only need to decide, “How can I optimize the number of orders today to do it the top way of efficiency with less cleaning, water consumption, energy? That’s where I want to be.” So to have that, we have seen several times where we have implemented that people are getting really happy, but you need to talk to them before and understand their needs because they are very important for that plant.

Mikkel Svold (28:13):
I’m imagining that just that simple exercise of controlling the priority of the orders to make it so water comes through this pipe and for the next batch it’s also water. So you don’t have to have that cleaning in between. Not that you would probably not need to clean that much if it’s water, but then say it’s raw milk. I imagine that in itself that could be the whole business case. Is that wrong or is it giving that too much credit?

Søndergaard (28:57):
You are absolutely right and when you look on the cleaning, it’s also a recipe how you can optimize the cleaning. You can think of, “Okay, what’s the energy consumption or the water consumption?” But some plants are only available to use a certain number of water so they cannot produce more. So what we see is with our engineer that they can optimize the use of water and chemicals. And have that in a recipe as well because sometimes you do too much to be certain. But I think Lars can tell much more about some of the projects. But it’s also the recipe instead of hard-coding, we know what actually to do.

Linnet (29:43):
But I also think you can say a lot of listeners here will know this requirement. But just for example, doing conventional and organic products. Sometimes you can go from organic to conventional products, but you can’t go the other way. Meaning that if you have produced organic material product on your line, then you can switch it to a conventional. And then that’s okay, but then you need to clean in between. And that is for example, one of the functionalities that can be really, really key to put in some kind of interlock or some kind of management layer and control system on top of that and can really provide not only value for the operator, but I also think that a lot of the end customers actually like the retailers, they also go in and ask questions from the producers, “How are you guaranteeing that this, this and this can happen? What is your procedure?” So that will also definitely be a use case.

Mikkel Svold (30:46):
So just here wrapping up lastly. You visit a lot of plants. Do you have a good example of here we could see this is a perfect quick win value right off the bat? Do you have a good example of that?

Linnet (31:05):
Well, I have one. For example, we had a producer when he had to order raw material, he actually had to go out the office, go out and look at the tanks and look at the physical instrument for how many tons was in the silo. And then he went back to the office and ordered. We did a project, we did this vertical integration from the shop floor all the way up to a management system and all of a sudden he could daily get a report of what was the raw material content of a silo. So he didn’t have to do that. So that’s just a very small everyday quality of life improvement that they got.

Søndergaard (31:48):
And the other area is the packaging line where you have the stops and said, “Okay, which one is the most critical one?” And start with that. And then when you have that up running, then you can build in more features. But I think the first thing is where’s the bottlenecks? Where is something that goes wrong? How can it be monitored or be fixed? So it’s start small and get a quick win. And then with the change management, if there come a quick win then more people is interested. So from the people perspective, a small step. But also from the risk, how many money do we use? If you take a small step and get value, then take the next step.

Mikkel Svold (32:39):
So instead of putting all your eggs in one basket, you would basically take them egg by egg.

Søndergaard (32:48):
Yes. Small steps. You learn each other. Because if I had a meeting with you one hour, I don’t know exactly the same. Where if you have been working for 15 years. And of course you need to learn a new product. So the more you learn from each other the better solutions.

Mikkel Svold (33:07):
Yeah. Erik Søndergaard and Lars Linnet, thank you so much for joining this talk today. Always very interesting just be talking about how you can actually implement data-driven decision making without anyone feeling run over. I think that in itself is a brilliant talk. Thank you so much for joining us today.

(33:37):
And to the listeners out there, if you enjoyed this podcast episode, do give it a like and share it with your friends or colleagues if they might find it interesting too. And of course also hit that subscribe button. It really helps us keep our spirits up and keep this podcast running and keep us making it happen. Lastly, I want to ask you if you have any questions, or anyone you think we should talk to on this podcast, or if you have anything in mind that good topic, something like that, do reach out to us at podcast@au2mate.com. I think that’s all I need to say. And I hope that we will hear from you guys out there. It’s always nice. That’s it. Thank you so much for listening.