Ep#152 Mastering Cloud Financial Management with Dieter Matzion at Roku

Host: Jon

Hi everybody, and welcome to the Faces of #FinOps, powered by ProsperOps. I'm your host, Jon Myer.

Faces of #FinOps podcast is about highlighting thought leaders in the cloud, financial management space, and insights and how they're making an impact not only within their organization but in the broader #FinOps community. Today's guest is Dieter Matzion, a senior cloud governance engineer at Roku, and also he's a #FinOps ambassador. Dieter is a member of Roku's Cloud technology and infrastructure team and supports cloud finance across AWS, GCP, and Azure. Please join me in welcoming Dieter to the show. Dieter. Thanks for joining me.

Guest: Dieter

Hey everyone.

Host: Jon

Dieter, I'm going to kick things off. Tell us about yourself and where you're from.

Guest: Dieter

I'm originally from Germany. I came on an internship in the United States and then got married and decided to stay.

Host: Jon

You came on an internship. What were you doing for the internship

Guest: Dieter

At university? When I graduated, one of my professors told me that out of all students over the past five years, I graduated with the highest score in databases. Was a surprise to me, but the professor suggested I should be using that for my career. So early in my career, I started building databases.

Host: Jon

Let's kick things off a little bit more and talk about your career and your journey to #FinOps.

Guest: Dieter

Yeah, of course. So eventually every company needs at least one database. So what happens is that once the database is built, I need to tune it. Once it's productionized and the transactions per second are going up, it becomes more of a tuning effort. So what I ended up doing is I ended up specializing in tuning databases and at some point I joined PayPal. PayPal. At that time, 2006 had the highest transaction database on the planet Oracle database, with 50,000 transactions per second running on a Sun E 15,000. So that was a big database. However, at PayPal, I also had the opportunity to. Then I stayed there for seven years. I had the opportunity to switch and manage the network operation center where I learned all of the operations of the different systems and subsystems, which was very interesting and exciting to me and a step to get out of the database area.

Host: Jon

So from PayPal, what were you doing next?

Guest: Dieter

Well, I sent a resume to Google. I was thinking, what's the worst that can happen? They reject me and then I'm no worse off really. However, Google came back and said, Hey, we are looking for someone who understands a complex environment with different systems and subsystems and can identify inefficiencies and carve these out so that we can move it to the public cloud offering. Google's public cloud back in 2013 was growing faster than it could build data centers. Building a data center is usually like a two-year affair, and while they had multiple sites that they were building the demand for Google's public cloud was so strong that they needed to find a stop-gap measure to use internal resources and assign it to the public cloud.

Host: Jon

Peter, I'm noticing a trend now. You went from databases, now you're trying to understand complex environments and Google Cloud. What was after Google?

Guest: Dieter

Well, as I was working at Google and doing great things, I noticed that I was only familiar with Google Technologies, and so I sort of became unemployable outside of Google. No one has Google, only Google has Google. So I was happy when Netflix reached out to me and said, Hey, that was in 2015. We are spending $250 million annually on Amazon Web services and we don't believe we need all of this, right? We need someone like you who can find these inefficiencies and tell us where we can save money.

Host: Jon

So Dieter, you went from understanding databases to Google Cloud to understanding complex environments and broadening your knowledge around #FinOps and trying to reduce the expenditure within environments, and that's where you went to Netflix, dealt with a w s. What's your next role? What are you currently doing now?

Guest: Dieter

Well, there was one step in between at Netflix if I found something, it didn't matter if it was a small dollar amount or a large dollar amount. Sometimes the engineers acted on it, sometimes they didn't. It was independent of the dollar size. So that was a little bit not satisfying for me that I do work and the work is not being acted upon. So I was thinking there should be a better way of managing that, like an organization-wide program where I then would go into the different business units and find optimization opportunities there I looked around and by accident, I found a job posting from Intuit and Intuit. The job posting was exactly what I just described word for word. The hiring manager asked me, why did you pick us? And I explained to them that your job posting was word for word what I envisioned.

Guest: Dieter

So at Intuit for four years I was running the back we call the cloud financial management program starting in 2016. The very first thing that people were saying was that they expected that there was a lot of waste back then Flexera came out with 30% of waste across all of cloud resources. So I started working on that. I built a system where we had waste sensors. A waste sensor is just a small little script that finds something and puts it into a database, right? Then we built a dashboard on top of that data that showed waste by engineering leader sorted by the highest dollar amount on top, and with six waste sensors, we found about 28% of waste. That was the start, right? Over the four years, we were able to improve that. We increased the amount of waste sensors. So we sort of raised the bar a little bit from six to 32, something like that, and we were able to push waste down from 28% to 5%. So that was great.

Host: Jon

So you took your knowledge from the database starting, and then you ultimately applied that to your, I want to say cloud financial management or your #FinOps. Was #FinOps utilized and was it an actual culture established or is it something that you guys were underneath the cloud financial management and eventually identified as a #FinOps practice?

Guest: Dieter

Yeah, so #FinOps existed back in 2016. I met J.R., Stroment at re: invent that year and he asked me, Dieter, do you think this cloud financial management thing has legs? Is that a thing? And I told him from my experience, as long as engineers are working, there will be waste. This is autologous to an industrial kitchen where you have multiple cooks you need to clean every day because the place will get messy. And I think that impressed JR quite a bit. And back then he was working for Cloudability which was later acquired by Apptio, and I think he was the Cloudability, he was the CTO, maybe even the CEO. And at Apptio he was the CTO for I think a couple of weeks, and then they let him go and in 2019 he founded the #FinOps Foundation. And so #FinOps is the first time then where this #FinOps term was coined.

Host: Jon

So Dieter, what are you doing now though?

Guest: Dieter

So at Intuit what I was doing is I did that waste management and then there's something called convertible RI’s. And so every week we would look at what is unused and we would use the convertible RIS to convert it to something that will be used in the next week. And that was sort of like an ongoing basis, but I also noticed that the forecasting of cloud spend was a huge effort, something around 30 people were working on that for the better part of three months, and the forecast variance at best was 15%, and at worst it was 70%, which is unusable. So I was thinking back then machine learning was a thing and I installed a machine learning system on my laptop, connected my laptop to the cost and usage report from a W s, and I ran the default machine learning model that came with the free software package.

Guest: Dieter

It was called a bag of neural networks, and it was able to do a forecast with an average variance between six and 9%. So I went to leadership and I explained to them that, look, I did this on my laptop, we can do something similar here for the forecast. 30 people over three months is a little bit excessive every year. Maybe there is something that we can do better. Now, while leadership was conscious that software can solve, there are forecasting problems. They didn't want to use what I had on my laptop which was sort of not an enterprise solution for them. So I was a little bit frustrated and a former PayPal coworker came to me. I was posting on LinkedIn and those kinds of things, and the person said, Hey, RO Roku is looking to build a cloud forecasting system. So I joined Roku during the pandemic in 2020 and started building a forecasting system that has a 2% forecast variance, which is fantastic. This is like an industry-leading thing, and so I was really happy about that

Host: Jon

2%. That's cool. So you were the first #FinOps hire at Roku?

Guest: Dieter

That's correct, yes. And the only one, we have a data scientist that helps me with reporting, but other than that, I'm the only #FinOps person there. Well,

Host: Jon

Let me ask you as the first thing that you did when you joined Roku as the first #FinOps hire, what were the number three things that you took on or wanted to achieve?

Guest: Dieter

Yeah, so I looked at waste first because that was my bread and butter and ruckus waste is less than 1%. So I was stunned when I found out the engineers are disciplined. They're turning off their workloads over the weekends and things like that, so there was nothing to do for me there. I was done. So while I was building the forecasting system, I also looked at, but in 2019, AWS introduced savings plans and at Intuit we converted all of our eyes to savings plans, which also means that I didn't have anything to do after that and the waste was already at 5%. So at Roku, one of the first things that I looked at is how the savings plans look like. And they were non-existent, so I built 80% savings plan coverage, which saves about 14% of the total cloud bill

Host: Jon

You came in and they had less than 1% of waste. I feel like that's a job already completed in a task. I mean, that's a hell of a discipline for a company to have 1% less waste.

Guest: Dieter

Frankly, I was stunned, right? Because I wanted to bring my waste dashboards, my waste sensors set everything up and then less than 1%, right? So I was a little bit disappointed, to be honest, but nothing to do there. I can focus on other things.

Host: Jon

Diet. How would you rank the company's maturity for #FinOps? Are you in a crawl, a walk, or a run using various maturity levels?

Guest: Dieter

It depends on the different #FinOps capabilities, and it's not wrong if you are in a crawl capability. Let's say you are a small shop, you spend like 10 million annually on cloud and you do your forecast in a spreadsheet, that's okay. You don't need huge automation there. You don't need to be the most sophisticated system Once every three or four years you do a contract renegotiation with the cloud service provider and you need to forecast what the trajectory will look like four years into the future. If you do have your spreadsheets, that's completely fine. Roku itself in the various #FinOps capabilities, mostly we are run, but I think it's also a culture for #FinOps to work, the executives need to know that cloud, as long as you build things in the cloud, there will be waste and they need to know that this waste cleanup, this #FinOps discipline is something that needs to be done on an ongoing basis. You don't just do #FinOps once a quarter and then you do something else. It's a continuous thing, right? There are cost anomalies, all kinds of training and credits, bill inconsistencies, invoices, wrong numbers on the invoice, and that kind of stuff. So it's an ongoing discipline. I have been doing it for 10 years now, so I can do a lot of things in very little time and that helps me get through my workday.

Host: Jon

When you say it's an ongoing thing, and I think here's the thing with the #FinOps culture and that it should be known is that, and I like that you pointed out that it's not done once a month or once every couple months. It's something that should be done. Do you recommend, obviously daily you're looking at the reporting daily, you're looking at anomaly daily, you're looking at cost either up or down daily as part of it so you can be in and stay on top of things?

Guest: Dieter

That's right. I have a daily routine. I look at cost anomalies in the morning. Some of those are self-service where the engineers get the messages. Some of us don't have the automation yet, and I just sent an email over with a screenshot of the cost graph It goes up, that kind of stuff, right? Then I look at the reservation management, and the savings plans, and our I portfolio improves gross margin by about 1.6%, so that is very substantial. If some random guy like me can help you save millions, why wouldn't you want to try that?

Host: Jon

Peter, what are some of the common or biggest mistakes that you might see from an immature #FinOps culture?

Guest: Dieter

It is, I think the most profound thing is to not know what #FinOps is and not know that it's ongoing. It's a continuous effort and not having that leadership support because it trickles down. I do monthly tech talks on specific technology subjects for both AWS and GCP where we train the engineers in new technologies, large language models, those kinds of things that are relatively new, but also older technologies where engineers may have made a mistake. For example, if you move billions of objects in an AWS S3 from one storage tier to another storage tier, there is a transition fee and it's like 3 cents for a thousand objects, but if you have a billion, that can be tens of thousands of dollars. So we may need to do a deep dive into that technology and then explain that a little bit more.

Guest: Dieter

And these technologies are different between the cloud providers. So for example, if you do go into GCP blob storage retrieval is instantaneous. You just have to pay a price for it. While on AWS, the rehydration takes time and the objects are being put into a staging area out of which they are being deleted 30 days later. So there is sufficient difference between those different technologies where the engineers might get confused and you need to provide training, but any training that I provide, if there is no leadership support, then I'm just spinning my wheels. I can't be effective.

Host: Jon

Leadership support is key for any type of implementation including #FinOps. You've indicated that this is very like a requirement for an immature #FinOps organization. Have you seen any best practices that you might recommend for a mature #FinOps environment or organization that you think should be implemented that might not be thought of?

Guest: Dieter

What we did at Intuit that worked swell is to have sort of quarterly business reviews that are #FinOps specific and you get a CTO, CFO, someone like that. Depending on the size of the company, your CEO could be present as well, and it's sort of like a sprint review meeting. This is what we did in the past, this is what we are planning for the future. Here are the current roadblocks that we are having, and you do sort of like this kind of business rhythm. You establish that business rhythm where the executives then not just get informed, but over time they will reach out to you to get information to make better decisions.

Host: Jon

Dieter, you were talking about Intuit where you used machine learning. Are you using any type of AI or automation within Roku or suggestions that others could use for their #FinOps practice?

Guest: Dieter

AI now with chat GPT, it's exploding, right? I have seen examples where you have a video and something happens in a video. Let's say the video shows a supermarket, someone goes and drops a drink and someone else comes and slips on that, You can then ask the large language model what happened and they will say, Person A came in and spilled their drink, and person B slipped on it and possibly injured himself and emergency services should be contacted. Some capabilities did not exist a year ago and they can be used everywhere, not just in #FinOps, but there are multiple initiatives at Roku for different things. Think of the video recommendations, and what you should watch next. These types of technologies provide substantial significant changes to what we were able to do in the past versus what we can now do with these new technologies.

Host: Jon

Are you using some of this type of automation to achieve or make the daily reporting within #FinOps a little more expedited or to provide the correct reporting?

Guest: Dieter

It is really funny. Two years ago, data scientists were just playing with things and it was sort of a novelty. Fortunately, at Roku, I have the ability I do some engineering work to get done, I can just open a Jira ticket, which is just a ticketing system, and then the work will eventually get done. It has to be assigned to a sprint and so forth. So I asked for a cost anomaly detection on GCP. They don't have a solution for that yet. AWS has a native solution. GCP is working on something, but we need this right now. So I asked and the engineer just naturally went and used the machine learning model for anomaly detection. I didn't even expect that it is such a prolific technology now. That was the first go-to, I would've done something with it if it's 20% more than the median over the last 30 days or something simple, No, they just slapped a machine learning model on that problem and it does a good job.

Host: Jon

And let's talk more about some of the things that you and your team are responsible for at Roku.

Guest: Dieter

I do a lot of different things, right? Cost and anomaly detection is one of 'em RI management and they have an RI automation product, but there are some blind spots. So we have a weekly meeting where we go over those and see what we can improve and help also, not just us, but also help the vendor improve their product. Then anything that is contract negotiations, I take care of that end to end, right? I negotiated the contract with the vendor, I got the contract through legal review. I work with legal on that, and then I also get it signed by leadership. So leadership needs to be briefed about that. I even pay the invoices. I'm working in the central technology and infrastructure team and they have 12 vendors that need their invoices paid. So I take care of a lot of different things on a day-to-day basis.

Host: Jon

Dieter, what are some of the biggest challenges that you and the team are facing right now?

Guest: Dieter

Challenges. I think there are technology challenges and there are also people challenges. I think interactions with people are the most important, the most outcome, and the most challenging because you don't know what is the knowledge of that person, what is the skill level, and at what level you need to explain something, right? So I typically start with the basics and work my way up. We have sometimes situations where I ask for something and the person comes back and says no, and I'm like, that's not a sufficient result here. You need to provide a little bit of context as to, why you probably talked to a team in the background and deliberated on this topic, so provide a little bit of context around what constitutes that. No, and maybe you should also come up with alternatives. We have one or two alternatives that we can do rather than just saying no, and it's these types of people interactions that I think I'm still learning and I'm trying to get better at that because I think everything that you do is done for people and if the people support you, your life will be easier and I'm happy to partner with them.

Host: Jon

Dear. I agree with you. What are your feelings on accurate reporting capabilities to make the most efficient decision about acting within the organization?

Guest: Dieter

That's critical, right? And it also has to be timely, right? I see Roku has about 2,500 engineers and 4,600 employees. There's a lot of experimenting going on and the engineers need to know immediately what is the cost impact of an experiment. For example, at Intuit we had a project where you can scan a receipt and it'll find the total on the receipt and then add it into QuickBooks. Not every receipt looks the same, Some receipts if you buy raw materials like lumber or something like that, the structure on the receipt is a little bit different, so it's not a trivial problem to solve, but the initial version was something like $5,800 a day and that was good for a proof of concept, but it is very expensive. We were then later able to tune the machine learning model that was used for that and get it down to $80 a day, which is much more doable and much more financial, financially feasible.

Host: Jon

You remember that the #FinOps survey came out a couple of years in a row, and this last year there was some indicator that 30% of the effort is getting engineers to act. You have over 2000 plus engineers. Are you running into the same issue and how are you approaching it?

Guest: Dieter

It depends on how you structure that. If you have leadership support, it'll be much easier because, it also depends on how much training leadership has, right? And how big is the engineering team as well? At Intuit, I had a situation where one of my leaders wasn't performing well. They were not responding promptly, and when I had a one-on-one conversation with him, it turned out it was just him and another engineer. So the team had two people, so be aware of what your target is, but the culture is critical to have this #FinOps culture and be able to just go to an engineering leader, say, here's the problem, and they will delegate it to someone else. And that happens daily that I find something, Hey, you made a purchase here on a credit card. We don't do this. We want this purchase going for an invoice, and I help them through that process. You need that understanding the knowledge and the support of leadership, and then the engineers will follow. Engineers fundamentally want to optimize stuff. You need to make cost another dimension that they want to optimize.

Host: Jon

I like your approach to helping them act. You didn't just go and say, you didn't do this action or you need to do this action. You understand why they're not doing this action, and then you help them improve that action so that they're ultimately able to do it on their own and understand why and the importance of it.

Guest: Dieter

That's key to success because you don't know what their knowledge base is, you don't know where their skill level is. I mean, sometimes it happens that I come to someone and I start explaining from the basics and they say, these are already know all of that. Tell me something. What do I need to do here? What are my alternatives? Right? And that's fine, right? Then I know this person is already from the skillset, a little bit more knowledgeable and I can work on a different level of

Host: Jon

Them. Dear, what is some advice you might be able to give an immature or practice is just starting their #FinOps journey?

Guest: Dieter

Do something, don't hesitate. I see a lot where people are unsure. Maybe they are still early in their career in the #FinOps journey and they're unsure about things, but whatever you do, do it in a way where you start small so that you can fix something easily rather than going all in and then you bought like 10 million of RIS and it was the wrong purchase and now you need to figure out how to fix that. But start with something. Failing fast requires that you do things fast. So I talked to someone, I do a lot of outreach as a #FinOps ambassador and I talked to someone who wanted to migrate their data center workload to the cloud and they had a three-year migration plan. I'm like, no, no, don't do that because there will be failures and you don't want this failure to be in year three. Do something right away in the next three months and then see if it was successful, adjust, and then take the next bigger challenge until you complete it. Don't hesitate. You need to act quickly to be able to fail quickly. If you act slowly, then your failure will be much later, and maybe the only way to correct is to switch companies at that point.

Host: Jon

I like that advice. Well, I feel enough. I'm out of here. Let's just say that I have a mature #FinOps practice. Is there any advice that you're able to give me to help maybe improve some things or stuff that are lessons learned that you went through?

Guest: Dieter

Yeah, look at things that you do manually often. Let's say you do forecasts every month and you are doing it with spreadsheets, Yeah, your process is documented, and it's repeatable, but it's still a manual process. Look at those processes and try to automate 'em. For example, when we started with anomaly detection on a w Ss, we put it into a Slack channel, and then we invited people to that Slack channel so that they could solve those anomalies, they can look at what's going on, and then post replies, Hey, I'm looking at that. I started looking at that. We looked at that. We turned off this workload. Try to introduce it gradually. Automation over time. Reporting will be key because decisions will be made on good data and timely data and what good data is depends on each organization will have a different view when it comes to allocations to shared costs, and how you handle that. So there are some very mature #FinOps practices like Kim Weir at Target has a team of 20 plus people. A lot of them are engineers who work on reports because reports are the key to decision-making. So from my experience, as you mature, there will be a higher focus on data within your organization.

Host: Jon

Good data, mature data, correct data. And the last few people that I've talked to for the Faces and #FinOps podcast, it's all about accurate reporting and timely data. I think it's all key to performing an action and being able to perform the correct action.

Guest: Dieter

That's very right. Engineers run experiments and they want to know the outcome of that experiment. An experiment could be, that I have an existing workload on a technology and I introduced one change to it. I go maybe from an Intel X 86 processor to an arm-based GRAVITON processor, what is the difference? They need this information immediately because when they do such a shift, what happens is that our eyes are being left stranded on the Intel X 86 and being unused, and now they are on a new workload that doesn't have RIs yet reserved instances. So their cost went up and they're like, what happened? I thought we would be saving money. Well, you need to tell me that you switched instances so that I can adjust the RI's accordingly.

Host: Jon

That's some sound advice because not only do you have the Intel, you have the AWS Graviton, There's also an AMD processor out there that has the performance base behind that and some cost optimization. You have three different types and you have to make sure that the risks that you purchase match the instances that you're using and everybody has to be aware of those actions that are happening.

Guest: Dieter

And that's also a challenge when it comes to forecasting because an engineer may build a forecast model for their new workload. It doesn't exist yet. It's going to exist in the future, but then they forget some line items like data transfer, for example, is a common one. So if that model changes the engineer, I don't know anything about that. They started with 40 instances or for a certain size, now they want to go with 20 instances and the forecast model changes, they need to communicate those changes. Sometimes I hear people complaining about that. The #FinOps and the engineers, need to closely collaborate in finance as well.

Host: Jon

Dee, there's something that I just learned about you during our very first conversation is that you are a contributor to the Cloud #FinOps book. Can you tell us about some of your contributions to the book and why?

Guest: Dieter

That is very interesting because O'Reilly Media approached me and said that J.R. Storment and Mike Fuller of writing this book and they need someone to review the book. They have lots of people that review the book on grammar and syntax and that kind of stuff, but they needed someone to review the book for technical correctness and they asked me if I could help with that. And so I have no O'Reilly Media author profile Now as a result, because I went through the entire content of the book and it was a Google Doc and there were multiple editors, it gets messy pretty quickly. Each editor is a different color where they're contributing, and then I would do comments saying like, Hey, you know what? This graph needs to be improved. It looks like it's a screenshot from taking from somewhere else. Do we have the rights for this graph, this technical statement here? Do we have some data to back that up I frankly don't believe it's correct. Those kind of comments. And it was a process. It took multiple weeks to go over that, and I was working with Mike Fuller directly occasionally, where we went over some of these items just to make sure that the book had the highest quality.

Host: Jon

Well, thank you for your contribution to the book and for helping out with it. I'm glad there's another person. I've got some cool information for you. So you had some speaking sessions at Reinvent in 2016, 2018, 2019, and last year, correct?

Guest: Dieter

That's correct. We took a little bit of a break during the pandemic.

Host: Jon

Well, it just so turns out that when I looked up your speaking sessions, I attended the one in 2019, which I think was my second or third reinvent session. I knew I recognized the name somewhere, and when I looked it up and I looked up the past ones that I attended and the documents and the notes that I took on it, I was actually at one of your sessions.

Guest: Dieter

That's awesome. Glad I was able to help.

Host: Jon

It's a small world and then we're turning it around here a couple of years later, and I get to interview you on this awesome podcast. So thank you for putting on those sessions. Do you know if you have any coming here in 2023 or can't you share yet?

Guest: Dieter

No, no. I can share. There might be possibly two sessions. It depends, right? I move quickly. Not everyone does, right, so let's see what we can do. I'm planning on definitely attending Reinvent. I'm also working with Cloud Academy. They just approached me and said that they want some #FinOps courses in their portfolio, so I do some work there as well.

Host: Jon

Dieter, how about we switch gears a little bit and have a little bit of fun, I asked you some off-the-wall questions. What do you think?

Guest: Dieter

Sounds good.

Host: Jon

All right. I don't want to prepare you too much for them, but I want to just pick out a couple of questions that help the audience get to know you a little bit more. My first question for you is what was the last book you read and why?

Guest: Dieter

To be honest, I don't read books. When you look at something like Malcolm Gladwell Blink or any of his other books, right? The core is maybe a few pages. However, you can't just sell a pamphlet for $40, right? He has to write a 500-page book, but I don't feel like I want to read through this 500-page book to get to the core of his message. So what I do instead is if the author is speaking at the conference, I will watch that video instead and then get within 20 minutes what he's saying there. For example, the spaghetti sauce, you don't sell more spaghetti sauce by making the world's best spaghetti sauce. You sell more by making all different types of spaghetti sauce with mushrooms, with extra chunky, all that kind of stuff, right? That's very good learning, but I don't want to go and spend my time reading through a 500-page book to get that same learning.

Host: Jon

That's very fair. I agree with you. Sometimes it's easier to talk to the author, do a summary, or even get the cliff notes for it. I do read a little periodically, but I do listen to some of the stuff, why I'm doing it? The audiobooks are really good. Mind, my next question for you is where would you be right now if you didn't need to work at all?

Guest: Dieter

I would be here by now. I established a sufficient enough portfolio that I could retire now, but I love what I'm doing. I love my work and I enjoy it. I derive satisfaction from it. So if I didn't have a full-time job, I would probably do more consulting on a needed basis and then substitute it with some income from the portfolio, but I enjoy what I'm doing. There are many people, who do the job, but because they need the income for me, I do the job because I love it.

Host: Jon

That's a good perspective to take for things. Exactly. Why you should do a job is because you love it and the money will come with you. Dieter. My last question is who are some of the most influential practitioners in #FinOps?

Guest: Dieter

Everyone. The #FinOps ambassadors of course stand out because they are the ambassadors. They're the one's Thesp spokespeople, but I learn continuously. I do a lot of outreach. I probably talk to something like 50 different companies a year so about one a week. I work with startups in other countries as well, Israel, where they have a product and they want to use me as a sounding board. I do benchmarking sessions where we talk to a peer organization and we try to find out where they are in the #FinOps journey. We share where we are and are, share our mistakes, and we want to learn from their mistakes, what they did, and how they overcame them in this area. You can't just read a magazine and get your knowledge from that. I prefer to just talk to people and learn from their learnings.

Host: Jon

Well, Dieter, I'd like to thank you for joining us for this Faces in #FinOps podcast. Thank you so much.

Guest: Dieter

Of course. Happy to be here,

Host: Jon

Everybody, Dieter Matzion, who's a senior cloud governance engineer at Roku

Dieter. I enjoyed our conversation and this has been another awesome episode in discussion around faces and #FinOps, powered by our good friends at ProsperOps. Be sure to hit that, like subscribe in, notify, and check out our latest episodes on the new YouTube channel and the ProsperOps blog.