Mettler-Toledo International Inc. (MTD) Earnings Call Transcript & Summary

Right. Good morning, good afternoon, good evening, and welcome to the Mettler-Toledo [indiscernible] Global webinar. The title of today's event is digitalization and data management, increasing productivity in chemical development and we'll be focusing on the emerging digital change that is happening in the chemical and pharma industry. My name is Bartosz Koprowski, and I'm a market manager for Europe, located in Nanikon, Switzerland, and I support our sales team across the continent and our customers in digitalization activities. Before we start, I just would like to invite you to really check our webinar website. You'll find it on our website, mt.com under live events. We also have a brilliant selection of prerecorded talks in our on-demand library. And again, it's available on mt.com/[indiscernible] webinars or simply Google Mettler-Toledo on-demand webinars and click on the tab Automated Reactors. And just real quick, please allow me to tell you about the upcoming events since this is our first webinar of the year. The next webinar is on the 22nd of February, the pathways combining data-rich experimentation with modeling to drive productivity, and it's going to be hosted by a colleague, Charles Gordon from Scale-up Systems. And then on the 29th of March, we have a webinar about chemical process research and development, hosted by Brian Wittkamp. The time is exactly the same as today. So it's 9 a.m. Eastern Southern Time U.S. or 2 p.m. GMT or 3 p.m. Central European Time. So thank you for joining us today. We've had a great number of registrations to this event. And you can see that it's clearly underlying the interest in digital transformation in the industry. Before we continue, I just would like to share a few ground rules. The presentations today have been prerecorded, and we did that in order to retain the optimal sound quality and also to simplify the logistics, but most importantly, we allow us to keep it to [indiscernible] talk around 30 minutes long. Please feel free to submit the questions in the MS Teams chat box during the presentations of [ relevant ] speakers or during the Q&A session. So after each presentation, we're going to have a little bit of time discuss the questions to talk to the speaker's life. And then the presenters for that will come off mute to answer your questions. We'll try to have it between 5 to 10 minutes each. For your information, this is an MS Teams live event. And in this case, the audio and video for all attendees have been disabled. And we did it to maintain the bandwidth and to minimize [indiscernible], but also our main task here is to make sure that we have the best sound quality ever. And again, please enter your questions or topics for discussions in the chat box at any time. Our first speaker today is Dietmar Flubacher from Novartis in Basel in Switzerland. So welcome Switzerland. Dietmar is working as a chemist in Chemical Process Development, so-called CHAD, within Novartis Pharma. He joined the company in 2009. And in addition to his work as chemist, he is responsible for lab automation and the electronic notebook. In combination to all of these roles, he's also interested in digitalization of the workflows related to lab environment, not only chemical but also for all functions within the technical development. The title of his talk is digitalization of lab data-solutions and challenges. And we're looking forward to the solutions. That will be all for me for the time being. Thank you, and please enjoy the talk from Dietmar.

So today, I will give you an oversight about our efforts on digitalization of lab data, the solutions that we found and the challenges that we also observed during this process. The agenda of this presentation is, first of all, I would like to share with you our holistic view on lab data. In the next step, we will have a small look to the standardization by using recipes, for example, using the iControl software. And then as a third part of the presentation, the look and feel on how we can access our data created on a daily routine. But let's start with the holistic view on lab data. As all of you know, we are creating lots of data with various instruments and equipment in our labs independent. If it's a balance, if it's a reactor system in our pilot plants, the reactor systems, for example, the EasyMax by Mettler-Toledo or any HPLCs that is being used in our labs. So all of these devices, they create a huge amount of raw data and metadata. And these data go into our documentation systems, documentation systems, for example, like our ELNs, for example that we all have or our LIMS systems that are also used throughout all organizations as far as I see, but also specific applications like Luminata that are dedicated to keep track of impurities, their [indiscernible], when do they get created, how do they get depleted. So with all these applications here, we are merging data coming from our reactor systems, for example, with data from balances or HPLC data here. All of our technicians make the connection between the real measurements, making the connections of data that belong together. So here, we are creating data with context. And these data with context, they can either be used for scientific interpretation and decision-making using applications, for example, like Spotfire, [ Qlik ] or any other BI tool. But this data can also go into submission documents that are used for the FDA or the EMA so that we finally can get our drugs approved. But how does that connect and how does that relate to an IT environment? And this is something that I can show you, hopefully, on the next slide. This is the setup that we are currently using within Novartis. So down here, we have our reactor systems. We are using [indiscernible], EasyMax, [ Chemspeed ] or even the DeltaV system from the pilot plant, they create a huge amount of time-based data. These time-based data is then stored and pushed to [ OSI PI ] systems. And here, we have a clear separation between the non-GMP lab data and the GMP environment where the pilot plant is storing its data. Any data that is stored on an [ OSI PI ] System is then made available via a virtualization layer. We call it DV12n, data virtualization one-to-end, meaning it is made available for any other application that can connect towards the data virtualization layer. So we are stepping away from point-to-point connectivities. We have one single source in case that you want to access the data, and this is the virtualization there. From here on, for example, with that data stream, dark green, the data is then pushed forward into our ELN and LIMS systems, for example, meaning we can pull our process data into our documentation systems, meaning we also are in control wherein case an experiment is screwed up, the reactor gets broken or whatever. And just repeat this experiment, the technician in the lab can decide which is the correct experiment, which data needs to be documented. Here, we also bring in the analytical data coming from [indiscernible]. At the moment, we are not in a full digital way, but we are improving on that part. The data journey doesn't stop here. For example, coming -- any data coming out from these data is then made available again via the virtualization layer, and then it can be pulled either into visualization and advanced analytical tools like Spotfire or [ R Shiny ] apps or whatever other tool our data scientists would like to use for their [indiscernible]. At the same time, the identical tables then can be -- are available for a data lake, as we call it, [ spine ] in our Novartis environment or it can also be made available for new applications like Composer or Luminata, these are applications that are in the POC state or currently being rolled out within our organizations. At the same time, here you can see a second virtualization layer. This is what we call the LabConnect, meaning making all data available throughout all the labs. This includes single devices like any titration, dissolution rates, balances and so on, all these data, they also need to be made available. But here, DV12n is not the right choice because DV12n needs database structures underneath for making the data available to anything else. And here, other applications, what we call LabConnect come into place, and they will fill up the gap if we want to connect all our devices into our e-notebooks or any other application that is needed throughout the entire system. What you see also here, the green ones are the data streams that are fully rolled out and the blue ones are the ones that are in the status of a POC or potential future connectivities that we see at the moment coming up. What does that mean for the technicians in the lab? Do they really have to take care? Do they really have to connect to DV12n? And I think here, you see this is not the case. Only the ones that are here framed in this orange box, these are the applications that the end users, the technicians, the chemists or the pharmacists, these are the applications that they are exposed to, meaning the [indiscernible], EasyMax, [ Chemspeed ] software, new applications or the visualizations in case we want to analyze our data, Spotfire, [ R Shiny ] apps and our documentation systems. These are the applications where the end users are exposed to. This is where they have interactions with. And everything else, DV12n, [ OSI PI ] systems, LabConnect, these are applications that are running completely in the background. They do the magic for the technicians. But one critical part is the standardization and our approach for the standardization of naming conventions is the use of recipes, as you can see in the upcoming slides. Let's take a step back, standardization, why do we go in that direction. So let's have a look to what we are currently doing as long as we are not in a digital world. So let's focus on one certain step of a chemical reaction. Let's say, this is the reaction completion after a certain reagent has been added. So currently, the situation will look like that our technicians did a documentation on perceived values, for example, the starting point of the temperature when the addition was done and then also in a manual way, the duration of the addition. Again, it's a very manual way and it's based on assumptions and prior knowledge because they do the documentation of the temperature, but they do not see if there is an exotherm during that reaction completion or if something else like the starting temperature, meaning, for example, the average or the maximum or the minimum of the inner-outer temperature is of relevance for further analysis. In case that this data would be of interest, this would mean that one has to go back, reanalyze all the data and do the documentation again. And this is something that hardly anyone is doing because it's a very tedious and very inefficient way to keep track of our data. So our idea was we take the same piece of data for the outer temperature and the inner temperature, we analyze this slot of data, and we directly provide the average, maximum, minimum temperature for the reactor and for the jacket. And in addition, we also document the duration of that step. So this also means that in the future, hopefully, no reanalysis of any experiment is required. In case that we decide additional values become of interest, then we just have to modify this in one of our systems and the data becomes available for all the experiments that I have been investigated. But how does that look like in the reality of a lab? And here, I think most of you are familiar with the iControl software. This is what the technician has to do. He or she has to design an experiment. You know that pop up, you have to enter certain metadata, and you have to use a template, which is standard template for any process step of a certain project. You need to check if certain tax assets, click on okay. And the entire recipe is already loaded for you, including the naming conventions for the different phases. And by using a recipe, you make sure that each phase of various experiment has the same naming convention. Now the experiment is being run -- sorry for that. Now the experiment is being run. And by the end of the experiment, you see these nice trends. And once you're done, you stop the experiment, you can add a comment, if needed, you click on okay. And then the magic is happening in the background. This data, as all of you know, is being pushed forward by the iControl software, pushing the data to the iC Data Center. This is a little bit tedious, but the amount of data that is being pushed forward is also quite big. And now the data is being pushed to the iC Data Center from where the [ OSI PI ] systems are picking up the data and they continue doing the analysis of the database on the recipe that was provided. So coming to the next part. Now the magic is happening in the background and what do the technicians in the lab have to do to pull to pull in the process data. So having a recipe, you take the idea of the recipe, you add a collection that is meant for documentation of process parameters to come up with the same name. And all you have to do is press one button, now the system goes, checks the [ OSI PI ] systems, checks the [ OSI PI ] systems, what are the data that are available for this experiment. It takes a little, you have to be patient. But I think it's worth the effort, specifically, if you see the amount of data, all the temperatures are calculated values. And I'm sure you will not expect any technicians to come up with the documentation of, for example, 67 temperatures or 12 steerings, beads based on what is done, including the times and so on. So now the data is stored. It is made available again via the virtualization layer. And in the next step, I will show you a little bit on how the look and feel looks like in accessing the data. This is the standard visualization. It's a starting point. It still can be modified based on the requirements, based on the questions that you have to your data, and let's have a look on how this looks like. On the landing page, I define the project. And based on that, I'm pulling in the data of all the experiments for this project. Let's say, I'm interested in the Step [ D10 ]. This is an internal naming conventions. Here are all the experiments, I can select them. I definitely have an overview on the purity, yield correlation. I can also modify that. I can, for example, could come up with a setup based on the creation date. And let's say, I'm interested in the times, the impact of various times on to any analytical data. This is what you can see here on this slide. So I made the selection of the experiments. Now I'm pulling in the data of the samples. I'm pulling in the times. And here, you can already see all the various times that are available and the range that have been investigated throughout these experiments. For the analytical data, again, you just click one button, and you have all the HPLC data in there. Now I'm limiting the data to the destination [indiscernible]. And here, I have the oversight of that correlation. One click, I can switch over to a different temperature of the process or to a different impurity and I can nicely see all the correlations that relate to that process. You can also do the same thing for the analytes. For example, you can bring in the analytical data, the metal trace analysis, all the palladium values for certain temperature. And here, you see the correlation and you can come up with scientific conclusions if there is a relevant effect, yes or no. As you can see, with the IT architecture, the use of recipes, we are in a position where we can really easily access all our analytical data and combine that with the process data that has been documented in our documentation systems. However, there was one unexpected challenge. All the things that are -- that you've seen here in this presentation, they are fully implemented. They can be fully used. But the tricky part is not the technology any longer, it's about change management. It's really about the change. And we, as a chemist, we are not native to data. Based on our training and our education and our experience, we still prefer running an experiment. But I promise you, we keep working on it to keep our chemists and our technicians trained that they become more native to data, and they really take a look to their data before they run the experiments. Finally, I wanted to say thanks to everyone who contributed. This is the entire lab automation network, [indiscernible], Michael, Pascal, Daniel [indiscernible], my sponsor, Daniel [ Kaufman ], IT support from [indiscernible] and thanks to all of you for listening, and I'm more than happy to answer any questions that you have.

Brilliant. Thank you, Dietmar, for this contribution. So ladies and gentlemen, please use the chart in your MS Teams live event to ask any questions to Dietmar. Right now, Dietmar is live with us, and he's unmuted. So if you have any questions, please don't be shy. This is your chance to discuss this presentation and ask Dietmar for his tips and tricks, experiences and stories. So before we jump in, I just wanted to ask Dietmar, I mean this is my private question. First of all, I comment super impressive progress. I'm following the development of Novartis since your last presentation almost 3 years ago, that was done in German. So I noticed the lab connected a new addition to the system. This is really brilliant. I just wanted to ask you, how does it -- I mean you have all these data streams, do you collect data just from Switzerland or just from Basel? Or do you also use other sites as well?

Interesting question. Let's put it this way, Bartosz. At the moment, yes, at the moment, we are collecting data from Basel and from China on one system. However, based on the requirements by the Chinese security law or cybersecurity law in China, we put some thinking in that if this can remain the way it is or if we have to modify that. It's really more about the assessment? Is this considered critical? What can be leveraged from that kind of data. So it's kind of risk management when you approach this concept?

Understood. Right. So -- but in this case, your colleagues in China, well, at least theoretically, have access to exactly the same services as the team in Basel in Switzerland. Is this correct?

They have access to the same services. They have access to the same infrastructure in the background. However, this does not necessarily mean that they have access to the same data. So based on the security roles, we can keep track on who can access which data in principle. If you need it, we can open the border. So the China colleagues can see data coming from Basel, but only on an as-needed basis at the moment. But the infrastructure from the setup on the documentation part, this is identical at the moment.

Okay. Good. Okay. We have -- well, we have 6 questions coming up in the chart. So let's just jump right in. Question from [ Colin ], very impressive work. What was the biggest challenge when integrating data streams from the lab [ OSI PI ] server all the way to the visualization software?

The biggest challenge there was answering the questions by the chemists. So the first time when I went out to the chemists and asked them what do you want to get? The answer was quite obvious everything. And then my next question was, what does it mean everything? And then you saw the reactions by the chemists, everything. Okay, please define it. So from [indiscernible], you have to bring in a little bit of structure into the entire discussion. How do you want to approach that? So this was one of the first challenges. Then the second one, this was a time when we originally started with [ pure SQL ] technology. And in the meantime, the architecture of virtualization layers came in, and that was a challenge for me personally, stepping away from -- I know that it can be done with [ SQL ] technology being open to something new. In the beginning, I [indiscernible] a little bit reluctant to use it. But in the meantime, I'm very glad, and I'm very happy that I listened to my colleagues in IT that they wanted to help me on that part. Nowadays, I'm not the perfect user on the typical data virtualization layer. But I can use it, and I can do many things in there, maybe not in a perfect way, but I can do it and I can provide benefit to the overall organization. And the last and the biggest surprise -- and the biggest challenge is what I mentioned in the end. So we have it up and running. It can be used. But convincing the standard chemist to use it on a daily routine. This is the main challenge. And I have to say I completely underestimated that part. And I hear that reaction from many colleagues within our organization. I'm not sure about all the other organizations out there. Maybe you are more open. You have different mindset by your colleagues. But I see this is one of the main challenges. We, as a chemist, most of us know it from our PhD thesis, I have a problem, I run an experiment, I have the results. So we are not used to -- let's have a look to what I already have. I have -- it's not our reflex on how to work with that. I think this is one of the major challenges that we within Novartis and I assume the majority of all people out there will face.

Awesome. So let's just carry on with more questions. There's a question from [ Stacy ]. Can you clarify that you had the same recipe built in iControl and that is the same in your ELN?

I'm not sure if I understand that correct.

I assume maybe there's some sort of debugging. I think this is what [ Stacy ] is asking, [ Stacy ], is this correct? Do you want to check -- and there's a follow-up question from [ Stacy ]. Does pulling the data back into your ELN slow down your ELN?

Let me try to answer that as far as I understand the question. The use of a recipe means that you define -- you have a process -- we are doing process development. So we know our products that we have to synthesize. So we have a process. We can define the recipe, come up with a standardized naming convention as a template. Once you take that template, you still can modify it. You can extend the steerings at times, you can play around with the temperatures, but what is happening within the basic elemental steps of that process, this remains the same. And therefore, by coming up with a naming convention on the phases of the recipe, you bring in the chemical understanding the process -- understanding to the IT system and then the end user who is consuming the data. Therefore, you have to use kind of a template which can be reused, which is one of the biggest benefit I see using a recipe. So this is for the first question, [ Stacy ]. For the second part. Pulling the data back into our e-notebook, we are not pulling back the raw data. The raw data, the time-based data is something that hardly anyone is really interested in there. Usually, I've seen it not that often that chemists or engineers, they really compare time-based data. Time-based data is shifting because one time the reaction takes longer to go to completion, and then you don't have an overlay bringing analytical data at least, I cannot consume that data in such a way that I can come up with conclusion. So therefore, usually, we compare aggregated data, the average temperature of the reaction completion? Or maybe there was an exotherm, maybe this results in some impurity formation. So these are the data that we are looking at. And therefore, we are not pulling in the huge amount of data, but we pulled back the small amount of data from the aggregation and this definitely is not slowing down the system.

Right. Okay. We have another question from Justin. This is about time spent, how involved were you in setting up all of the data connectivity to your database. And then he's asking how many full-time equivalent would you estimate were involved, how much budget and how many years? Wow, [indiscernible] one of those killer questions. So do you want to elaborate on that? Thank you, Justin. This is a very good question.

Making the connections mainly, these are the connections between the [ OSI PI ] systems and the virtualization layers between e-notebook and the virtualization layers. If you have all the information available, the connection string, the passwords, the ports have to be open, then it's done within 5 minutes for each connection. It doesn't take much more. The challenge, and this depends really on each organization, how do you get the people together, having the knowledge who can provide the password in that instance, things like that. So I would say making the connection between system A and system B to the virtualization layer or the other way around, this takes about, I would say, 1 week until it's really fully implemented for each system development test and production environment. This is the time frame, hours really spend can be something between 1 hour and, let's say, 8 to 10 hours real working time. But this is more about getting the right people together. Over time, you get the experience and then these things are getting faster. Sorry, what was the second part?

Budget.

Budget. For all the things that you've seen here, creating the views in -- on the e-notebook level, pulling the views into the virtualization layer, creating the new views there, exposing it to Spotfire, building the visualizations. This -- what you've seen here was done by more or less one person, me. Time spent. I would say for everything that I did here would be about half a year full time. So this was spread about 1, 1.5 years. So this was about 50% of my time. The rates, how we are paid are different between each company, between each country and so on. So -- but based on that, I think you can backcalculate. And again, keep in mind, I am a chemist. I'm still working as a chemist. I still have process projects ongoing. And I'm doing this as part of my job. So I'm not 100% IT. I'm not 100% data scientist or whatsoever. I'm a chemist, I'm not the expert on all these tools. But even for me, I was able to pull up these things within a reasonable amount of time, I would say.

That's perfect. Cool. I think [indiscernible]. We're going to have 2 more minutes. There are a lot of questions, so I apologize for not reading all of them out as he is going to do some trials here. Right. Have you been able to integrate the data for [indiscernible] into modeling software, such as Dynochem?

So far, this was not the requirement, but the same approach here, Dynochem would pull up the data via the data virtualization layer. So I'm not providing any additional tables or whatsoever. Any tool would pull it in from there. This is also the same approach that we followed up with the data scientists. So they are not pulling data from e-notebook. I don't want to create additional views there. They pull whatever we have directly from the virtualization layer. The good news is this is not limited, the entire architecture that you've seen here is not limited to chemical development. We extended the entire use of that architecture also to our colleagues in formulation, and they are connecting the [indiscernible], which is like the technical plant, the handover between non-GMP and GMP, the first test on the first 1,000 tablets and so on. They connect their devices into the same architecture using [ OSI PI ], DV12n, e-notebook and so on. So it is really fully scalable to whatever we are doing out there in the pharma world.

Last 2 questions. What measures have you taken to make chemists and technicians more native to data?

We are working on it. I played around with that. What I see most at the moment, which is helpful is that you really have to go into the project teams, and you really have to discuss that in small groups and show them the benefit. At the moment, I'm working closely together with a technician. She is end 50s, let's put it this way. So definitely not one of the IT addicted guys out there. And from the beginning on, she realized that there is a big benefit in there, but she was a little bit scared about the flexibility, what do I see here when I pull all that together. And if you sit down and you help them and you can do that in smaller groups. It doesn't have to be one-to-one, it can be in smaller groups, project teams, something like that. This really helps people, and they get eager and curious to see and you will see some nice surprises. I had discussions with chemists working on oligonucleotides. And they -- during one of these small group discussions, they came up, but we are still facing issues with our -- with the stability of the analytical method, did you take a look to your data? Can I do that? And then I showed them within 2 minutes on how would this -- they were sitting there -- interesting, yes, yes. I learned something from here. So you will see nice surprises, and you will be able to answer questions that you feared to ask because you were scared that you have to pull all the data. But now with all the analytical data and they are important, this comes into place, and we have it, and this is also a good starting point. I would not start with all the process data because this is really high end. But what we are doing on a daily routine, we have to discuss analytical methods, independent if it's early phase or late phase development. So there, these tools also can help, but it's not the focus of this presentation with the process data coming in. But as a starting point, focus on analytical data because this is something that we are used to and that we have to discuss anyhow.

Okay. I really want to -- we need to push forward, Dietmar, really quickly, 15 seconds. Last question. So ladies and gents, please, we'll keep your questions in the chat. We try to address them, so we'll try to get Dietmar to answer them. If not, we'll try to email you the answer. But there's one, really cute one that I want to point out here, we are a very small company trying to implement exactly the process you are talking about. How did you get buying from management? How did you sell it to these guys?

Definitely, it was beneficial that our CEO was very fond of data. So this helped a lot. At the same time, just do the math. When I started this project, I really went into the labs, sat down together with a couple of chemists, a couple of technicians with a stopwatch, how much time does it take for you to put the data in there. And then when you compare that, what you can do afterwards, I'm pretty sure each and every manager will say, go ahead.

Perfect. That's brilliant. Fantastic. Right. So in that case, thank you very much for your time, Dietmar. I know that you have to go diving after that. So I hope you [indiscernible] brilliant. I hope you'll give us a few more moments with the questions. And if not, if we have your contact details, and we can identify you, we'll reply to you via e-mail. Again, ladies and gents, thank you very much for the questions, and we're going to move on now with the next speaker.

Right. So our next speaker today is Allyson McIntyre from AstraZeneca in Macclesfield in the U.K. Allyson is a principal scientist who is passionate about maximizing understanding of the chemical processes. Allyson is also excited about what can be achieved through the use of technology, process understanding tools, and data management. The title of her talk today is integration of data structure between iControl and electronic lab notebooks-AstraZeneca's journey. So thank you, and let's hear from Allyson. Ladies and gents, sorry. It seems that we're having some issues with presenting the right file. So my colleagues in the U.K. -- sorry, in the U.S. are addressing this issue. Yes, we've had a couple of problems that reported by Microsoft today and been happening all day. So that affects Office 365. It's kind of like an anti-ad for Microsoft services and also MS Teams. So it seems that we just got affected. But in the meantime, maybe we can do -- all right, okay. There might be some content coming up.

Hello, everyone. Thank you very much for joining us for the webinar. My name is Allyson McIntyre and I work in Early Chemical Development within AstraZeneca and the Process Development sections. Today, I'm going to talk to you about the integration of data structure between iControl and ELN and more specifically about AZ's journey in this process. In process development, we have the [indiscernible] Technologies network. This network is used to be able to help support our scientists to be able to complete their experiments at the right time using the right equipment. We use tools such as Lean and standardization, employ new digital technologies and techniques to be able to help support them through well-designed, high-quality data collection and easy-to-use technology. The hope is that scientists are able to use that data to increase their knowledge and understanding of our process. At AstraZeneca and the chemical development space, we have over 70 automated lab reactors now. This is comprised of EasyMaxs, OptiMaxs and RX-10s, which allow jacketed vessels to be used in an automated fashion. To complement this, we also have 13 autosamplers to enable automated sampling alongside our automated lab reactors. And we also have a fleet of process analytical technology instruments, including FBRMs, EasyViewers and [ PVMs ], which enable plug-and-play technology alongside our automated lab reactors to ensure that we get good quality data and understanding of our processes. So within the chemistry, we had a vision set out, which covered kind of 4 major areas. So we've got the LabConnect project, Project StepStone, [indiscernible] project and then the integration with ELN project. And each of these projects came around from a different kind of aspect of that vision that we wanted to try and achieve. So first of all, we want to be able to enable scientists to collect the process data from the experiments easily. So this was the LabConnect project, which was about bringing in the Mettler-Toledo fleet of equipment to enable our users to be able to use that to automatically do their experiments while collecting process data in the background kind of for free of charge, so to speak. The second part was around being able to enable that data to be collected in a more structured way. So this was a joint project between Mettler-Toledo, AstraZeneca and Janssen Pharmaceuticals, the [ J&J section ]. So we were looking at being able to collect that data in a structured way that we could then use to be able to kind of search and visualize in the future. And that project has been underway since 2015. And I've now recently just passed that over to a colleague at AstraZeneca to carry on with. And then the final component was really around having the scientists being able to search and then visualize that data. Now to enable us to be able to do that, we needed to be able to change and adopt the way that we integrated within AstraZeneca. So this actually borne out 2 different projects. We had the integration with [ ELN ] project, which allowed us a direct link with our ELN to be able to send information and retrieve information back in an automatic fashion to enable our scientists to have an easier workflow. And then the second one was the segregated network project, and this was around being able to have all of the software and the hardware on a single system that can communicate with each other in the correct way to enable all the data transfers to happen the way we needed them to. So what I will do today is I'll talk you through each of these 4 project stages to be able to enable us to get our lab chemistry vision and talk to you a little bit about the pros and cons and how we managed to achieve that. So fundamentally, what we're trying to do is we enable our scientists to collect good quality data to be able to interrogate that data and get greater understanding of the processes while being able to reduce nonvalue-adding activities and reduce the burden on writing up experiments. So fundamentally, what we're attempting to do is allow scientists to focus on the science while we enable the technology to do some of the hard graft for them in the background. So to start off, [indiscernible] discuss a bit more around the LabConnect project. So over the years, we've been able to change how we use our equipment and what equipment we do use in our labs to be able to make it easier for scientists to do the work, but also to get more understanding from each experiment that we do. When we started this journey, we use things like round-bottom flasks and [indiscernible] vessels with very basic temperature loggers. We then moved into some more kind of automated devices, but they tended to need specialist support to be able to run them. So they really didn't have a huge uptick in our labs. And then finally, around 2014, we started to investigate the use of automated lab reactors from Mettler-Toledo, and we ran a successful trial in 2014, where the chemists utilized a couple of bits of equipment from Mettler-Toledo to be able to use this for their project work. What they found was that it was really simple and easy-to-use software and hardware, which enable them really just get on with their job and not have to kind of think too much or too hard about it, and there was a large [indiscernible] of training for them as well. In the subsequent years, we've been able to work with the chemists and our business to be able to put cases together to be able to expand this. And we now have a fleet of over 70 pieces of equipment that our scientists have access to as well as the accessories that go alongside them. So the concept around LabConnect was to enable the scientists to be able to use the automated lab reactors to complete their experiment while collecting the process data in the background. So the combination of the ALR hardware and the software platforms using iControl and iC Data Center, we're able to automate part of the process of an experimental write-up and capture that data to provide a human-readable form for the scientists. The idea was that, that human-readable form would be able to be put into the ELN and form part of the write-up for our scientists reducing the burden and what they would have to write up to be able to complete their experiments. Over the years, we've been able to evolve the LabConnect system and automate parts of it to make it easier for our scientists to be able to do their work. However, there is still quite a lot of things that didn't really go the way we wanted them to go with the system, including a manual system to system transfer from our ELN to iControl and the LabConnect equipment. We had time-consuming upgrades for our specialists to be able to make sure we maintain the most up-to-date software and links that we needed. And then we also have things like transcription errors because the scientists had to manually transfer information from ELN to iControl, then this created some transcription errors. In addition to this, the way that our setup was due to security reasons, it also meant that we had to have iControl open in order for any data to be collected. And this meant that even for a scientist doing a really quick experiment, they weren't able to just use the functionality in the touchscreen. They have to make sure they logged onto PC and ensure that iControl is open and connected or they wouldn't retrieve any of their data. In addition to this, some of the structured data we collected was minimal and didn't provide as much information as we would hope to be able to gain understanding in the future. Once the experiment was complete, we did have more automation on the back end, which allowed us to kind of move the data from the system through iC Data Center, and then that gave us a hyperlink into our ELN and our data storage archive unit. However, the information still has to be manually transferred to the ELN so that word document kind of human-readable version that I spoke about previously. While originally, the aim was to have that automatically [indiscernible] ELN. At this point in time, we had to have that as a manual process. What we found was because it was a manual process, our chemists didn't do it and therefore, ended up actually repeating their work and writing out free text into their ELN. So this was causing a duplication of effort, which wasn't really something that we were hoping to get from the system. In addition to the kind of poor setup, we occasionally did suffer from data loss or inaccurate information where the chemists haven't necessarily put in all the correct information because they didn't want to duplicate all the work that they were doing. So this really brought the kind of hardware up to kind of business as usual, where we currently review the needs of the scientists and the advances in the technology with Mettler-Toledo to be able to provide the right technology for scientists. So on a yearly basis, we catch up with Mettler about the kind of new advances and with our scientists about the requirements that we need, and then we adapt and upscale and change our kind of hardware fleet accordingly at that point. The end-to-end infrastructure is lacking. It's not where we would like it to be, and there's a poor uptake of that human-readable report, and this is what really kind of strive us to do some of the other parts of the project to really try and improve this work. So while there might be some kind of downfalls to the way we have this set up, there were some real advantages noticed, especially during the COVID pandemic, with our users kind of struggling to get on-site and trying to limit the interactions we had, we were [ unable ] to actually take advantage of our system setup for LabConnect. All of our PCs were remotely accessible. So this enables our chemists to kind of log on remotely from home and start their planning in iControl before they even came on-site, which then essentially minimize our lab time required. So the scientists really only came on-site to be able to just set up the experiment, press play, and then they went home, and they could monitor it remotely and make small adjustments as required. And then the second kind of back end of the system because it was all fully automated, it meant that the user only really need to come on-site to set up and to clean up. Everything else, they were able to remotely do at home, which really reduce the kind of the requirement to be on-site. And we really found that that's where the kind of chemists started to gravitate towards. So rather than having to be on-site for too long, they tended to kind of use this equipment to be able to do it. So this is kind of really showing kind of here within the graph that I've kind of produced. So you can see from kind of 2019 to 2020, even though we were in a year of pandemic, we actually still managed to increase the number of experiments we were doing across the year. And we kind of held steady around the kind of initial lockdown periods. So that kind of red zone and into the kind of green zone, there was a lot of lockdowns in the northwest of the U.K. during COVID, but it didn't actually minimize the number of experiments that our scientists were able to do even though they were kind of not coming on-site as often. So we were very kind of lucky that we were able to get our scientists on-site. But as I said in the previous slide, they used to come on-site, set up their experiment and do as much as possible they could from home. So the way that we have the hardware setup, it really enabled the scientists to complete as much work as possible remotely and minimize that time frame to the on-site and ensure that we could still get kind of our reactions going and we didn't actually stop our work, which was really, really important. So the StepStone project was done in collaboration with Mettler-Toledo and J&J. This was a project to try and enable greater data structure within the Mettler-Toledo platform. So Project StepStone was around trying to kind of create that data quality through structure. So previous to kind of the implementation of Project StepStone, when we were using our ELN, we had a [ free-text ] approach. So this is unstructured data, [ per reuse ] of knowledge and information as it wasn't always available or even accurate, and we didn't really have any harmonization. So one user could write one thing and another user could write something differently. So what we wanted to do with Project StepStone is really bringing this structured recipe-based approach, which is around kind of having standardized structured data and the ability to have both human readable and machine-readable form that would be compatible with multiple sources and essentially giving us a way to have accurate information available for reuse. So we wanted to be able to do this in a way that it's captured automatically as the chemists worked. It wasn't necessarily having to be a kind of extra burden upon them when they were using it. So the [indiscernible] process model that we used has been developed and shown here. So we have kind of higher-level process, which describes kind of what kind of experiment is we're doing, then we've kind of got the stage. So each experiment can be built up from a number of stages, which really describes what the process of the experiment actually is. So it could be something like reaction and then extraction and filtration, for example. And then each stage is then composed of a number of operations, which really describes what it is you're actually doing within the experiments. And then under the operational level, we then have the parameter level. So every operation has a series of parameters that then describe that operation to be able to kind of give you more context and information around what actually happened. So the idea is that this is all embedded within the Mettler-Toledo iControl software. So it's kind of drag-and-drop technology or select within the touchscreen of the equipment. The user essentially is just operating the technology as they would have previously, and they get pop-up prompts to be able to fill in to kind of provide the extra contextual information. So the idea was to be able to generate high-quality data through this execution phase. So we wanted to really have as much automation there as possible. We wanted to have user-friendly software in a flexible working environment that was intuitive for our chemists, which would ultimately lead to accurate real-time recording of our structured data. So what we've done is we've implemented this with the Mettler-Toledo software platform. So either on the touchscreen or within the iControl platform, you now have simple, easy-to-use functions to be able to bring in this extra structured data. And then at the end, we have a human-readable report, so the user can be able to read through and understand and repeat the experiment. But we also then have a human -- sorry, a machine-readable format that we can then utilize to be able to search and visualize in the future. So you can see here as an example for our model that what a user might do is, they might select their process. So here, we've got synthesis. Then their stage, for example, filtration and then their operation filter. So what you see inside the software is you then get to pop-up box for the user as a prompt to be able to fill this in. They can fill this in at various stages either before, during the planning stage, during the running stage when the experiment is actually running or even they can edit some of the information afterwards in the analysis phase depending on what it is. When this is kind of edited and at the end of the experiment, this is all kind of then translated into both this machine readable and then the human-readable format. And the idea is that it's used by the chemists and by data scientists in the future to be able to do extra work with. So with Project StepStone, we've delivered the first version of the software with Mettler and J&J, and we're currently working on the next version upgrade, which is 6.2. So due to some of the kind of poor IT infrastructure that I mentioned earlier in the presentation and an ongoing internal ELN project, it's meant that AZ has actually been delayed in the implementation. So we haven't been able to implement 6.1 when it was first released. It's only more recently, and I'll get on to that with the rest of the talk as to when we've actually implemented 6.1. But 6.1 is there. It has got the [indiscernible] model in the background, and that functionality can be turned on. And if you are interested in finding out more about it, then please reach out [indiscernible] contacts. And if you want to find out a bit more about kind of how we've implemented it, then feel free to reach out to myself or I can pass you on to one of my colleagues to discuss it further. So one of the IT infrastructure projects that we had to undertake to be able to enable us to have all of the structured data and good connectivity with our LabConnect fleet was the segregated network project. So the aim of the segregated network project was to allow us to have the appropriate IT infrastructure to enable our lab connect ecosystem to actually exist and operate as we intended it to. So we needed this 3-way communication between IC data center, the LabConnect PC with iControl on it and the LabConnect ALR touch screen. . Due to security constraints within AstraZeneca, we weren't able to just plug our instruments directly into our network. So it meant that we had to create a separated network that had all the right infrastructure and connectivities across our network phases to be able to enable this to work. So the idea is that experiments can be carried out on the ALR without the requirement for PC, which is something that we didn't have in the past. All the data will be collected and stored in the lab, so we would have 100% of our data integrity. And then also, we would have a suitable link with our future ELN by enabling us to have this full communication and connectivity piece as well. . So this is just a kind of quick oversight of kind of what actually occurred as part of the segregated network project. So it's quite a complex diagram, but it just kind of shows all the kind of intricacies around actually trying to be able to create the correct communication. So on a kind of high-level simplified, you're just trying to create a 3-way communication. But in terms of an actual data workflow, there's a lot more kind of movements. So we've actually got about 17 data movements within this slide. And you can see from the colors, we're actually moving from one network to another. So we've been able to work with our IT partners to ensure we have the correct data flow going from one network to another network and having the right firewalls in place to make sure there's no risk to either our equipment or any of these edge networks. So here, you can see that our new ELN will be able to send -- automatically send information for new ELN into the LabConnect ecosystem. Then once it's in the LabConnect ecosystem, the data center picks that information up and can send it to our ALR and our iControl PC, where the user is able to carry out their experiment as they normally would. On completion of the experiment, that data is processed by IC data center and dropped into a series of folders, which enables the users to be able to access and execute -- access and added any of the data that's required. So we essentially have what we've termed as kind of a process folder, and then we have like what's the raw folder that the user doesn't have access to. That then enables us to kind of go up to our archive units so that we've got data or raw data stored and archived in the future. And then we also have a link back to our ELN so that we can automatically pull our actual information back into ELN as well. So with the segregated network project, we've successfully created that network and delivered it by IT earlier this year. We've been able to transfer 140 systems onto the segregated network. So that's all of our ALRs and their associated PCs all needed to be transferred onto the system. The LabConnect ecosystem is now fully supported by IT, and this was a big win for us because previous to that it was only supported by the Applied Technologies network, which meant if people were on holiday or wait conferences, for example, the chemist didn't have the support they required. So now it's fully supported by our IT department, so users can raise a request if anything goes wrong, that they need help and support with. It also enables -- it's an enabler for our ELN to iControl integration, which I'll talk about next. The iControl integration project with our ELN was the idea to be able to have a user able to create information in their ELN and then send that information into the lab into IControl. In the past, as we kind of evolve through our LabConnect processes, we found that users didn't like to have to duplicate their information, and we often got transcription errors when we duplicated that information. So we wanted a way that we could actually send that information to reduce the need to duplicate and also then reduce those transcription errors as well. So within the LabConnect ecosystem integration, we wanted to integrate several different properties. So the first was the chemistry table. So we wanted to continuously -- continue to use this synthetic table within the workbook. It's what our chemists are familiar with and they use routinely. And then we want to be able to send the planned amount through the IControl. And then we wanted the ability to update those actual amounts and any new materials from iControl. It's essentially given the flexibility to the scientists that they can kind of know what they want to use upfront, but it doesn't prevent them from changing their mind when they get in the lab if something happens to the chemistry that they require a new chemical or they need to change the amount, for example. And then we also wanted to be able to update our cost with any new materials because our SOP states that we have to kind of update the caution and everything else. So we wanted to try and automate as much of this as possible. So the idea is that the chemist goes into their ELN. They can start with their synthetic chemistry table and populate it knowing what chemistry materials that they need. They then send that directly into the kind of chemistry section of iControl. So all they need to do is press the button. That will automatically get sent through the background through that data flow that I showed you in the segregated network project and then picked up by IC data center and pushed down to the systems so that the chemist can use it. And then on return, again, by pressing button in ELN, they're able to bring that information back in and update any of the desired or changed information within ELN to look an up-to-date system. So essentially, you'll be able to get the data that you require. You're not having to transcribe it from one system to another. It's automatically moving between them, but you've still got the flexibility to kind of add new materials or change the amount as you require. So the next element for integration was the word report. So this is the automatic write-up of the procedure as a word file. So essentially what I've called the human readable version. So within the Mettler-Toledo system, it automatically produces a word file while the experiment is being processed, and we wanted that word file to be set straight to our ELN. So as I mentioned earlier in the presentation with our evolution, we found in the past when this was having to be a manual procedure, a chemist didn't do it and ended up rewriting everything in free text, while this wasn't helpful because we didn't have access to the human readable version that was automatically produced. We also then ended up duplicating our efforts and writing it in free text. So the idea was to bring the word file indirectly, and this enabled us to be able to edit and add further details, for example, in the analysis and diagrams for an example. So what we've been able to do is, again, when the user hits the bottom, it updates the chemistry table, but it also then brings in that word document and then they just right-click and edit it, open up in word and they have all the full functionality more to be able to then edit and add extra information into that. And then the final piece around the integration was around the data use. So we wanted all the data to be brought back into our ELN from iControl to enable us to use that data in tables and graphs and some of the data to be used fundamentally in calculations within our ELN. So again, when the user presses button, all the data is brought in, it's stored in kind of predesigned tables so that you can see all the information. As you can see, here, we've kind of got the different operations that we use throughout the process and then the associated parameters in the context of the parameters because this is all tabulated in a structured form, it means it's search in the future. So our data scientists can use it to try and interrogate and find out extra information as well. So the LabConnect ecosystem integration was really a true kind of 3-way process between BIOVIA, our chemistry department and our IT department with a lot of help and support from Mettler-Toledo in the background as well. We wanted to be able to have this kind of flexibility to be able to do what we wanted in the lab, but still have the ability to kind of structure that data and use the chemistry tools that we wanted to do. So by being able to do this, we've been able to incorporate the chemistry table and have an integration between our ELN and iControl. We've been able to pull back the word document after the experiment is completed as well as the data to be able to use that data in the future. Meanwhile, been able to use a kind of user friendly, easy-to-use software to be able to collect that data in a structured way that feels kind of natural and kind of part of their everyday life for experimental scientists. So at the moment, the integration with ELN is working and rolled out. So we rolled out in August this year. The users have been trained and are actively using the system with the data structure enabled. The connection to our ELN, there's quite a lot of training. So it's not necessarily on the Mettler-Toledo side. But the connection in our new ELN, there is quite complex systems. There's been a lot of training. So that has kind of led to kind of decrease in our productivity this year as we try to instill this. And currently, we are having an issue with the difficulty of the sizes of the data files. So because of the way that we're integrating with our ELN with the data side, this has caused some problems, and we've had to shut off one element of the data that we're bringing in. So we're still able to bring in the process data, so all the modeling and everything that I showed you there. But the trends data we've had to put a pause on while we investigate what else we can do with that because it's just causing performance. So we've had to switch that off. Now the one thing that we've been able to do here with bringing the data into our ELN is allowing us to have within experiment comparisons. So it allows us to kind of compare what's happening within that experiment and look at any differences there. But one of the things that we want to do in the future is to use this data reuse to give us greater insight, we want to be able to extract that process information from iControl, and we want to be able to combine it with our analytical information to provide greater knowledge for our scientists. So the idea is the ability to pull the basic information together, and we're doing some ongoing work being able to pull that data from our ELN into data lake to be able to then do between experiment comparisons as well in the future.

Right, ladies and gentlemen, it seems that we got a few issues with the last 2 slides. But maybe we can already jump straight into the Q&A section. Alison, I know the you are online, which is great. Thank you very much. That was a very impressive content. We really had a few questions, but I'd just like to ask basically, could you summarize the last thing because I think we had some sort of kind of strong hands on the each [indiscernible] comes to MS Teams.

Yes. So just to summarize at the end is basically just pulling everything together to try and connect everything up. And the next step is really about data visualization piece. So we've focused quite a lot of work about being able to collect that data and then structure that data. Then the next section is then really trying to retrieve that data back out and start to actually visualize it and reuse the data. So my colleague that's now taken over the project from me, is going to be working with some of our data scientists and trying to extract the data and convey that and see what we can do from some of the kind of data reuse side.